M

July 16, 2025 12:18 AM GMT

North America Insight

Global Fixed Income, Tech Diffusion, & GenAI

Bridging a $1.5tr Data Center

Financing Gap
We think credit markets – led by private but aided by corporate and securitized – will be
critical to meet $1tr+ of data center financing needs through 2028. We remain bullish on
the growth and return potential in private credit and data center ABS/CMBS. Corporate
credit should see manageable bond issuance skewed to higher quality. Risks to financing
capacity in credit are sharply lower rates/weaker growth, which could dampen end-investor
demand.

Morgan Stanley does and seeks to do business with companies covered in Morgan Stanley Research. As a result, investors should be aware that the firm may have a conflict of
interest that could affect the objectivity of Morgan Stanley Research. Investors should consider Morgan Stanley Research as only a single factor in making their investment
decision.
For analyst certification and other important disclosures, refer to the Disclosure Section, located at the end of this report.
M
North America Insight

Contributors

Morgan Stanley & Co. LLC Morgan Stanley & Co. LLC Morgan Stanley & Co. LLC

Vishwas Patkar Lindsay A Tyler Carolyn L Campbell

Strategist Credit Analyst Strategist
+1 212 761-8041 +1 212 761-2734 +1 212 761-3748
Vishwas.Patkar@morganstanley.com Lindsay.Tyler@morganstanley.com Carolyn.Campbell@morganstanley.com

Morgan Stanley & Co. LLC Morgan Stanley & Co. LLC Morgan Stanley & Co. LLC

Vishwanath Tirupattur Stephen C Byrd James Egan

Strategist Equity Strategist Strategist
+1 212 761-1043 +1 212 761-3865 +1 212 761-4715
Vishwanath.Tirupattur@morganstanley.com Stephen.Byrd@morganstanley.com James.F.Egan@MorganStanley.com

Morgan Stanley & Co. LLC Morgan Stanley & Co. LLC

Kyle B Yellin Catherine Liu

Credit Analyst Strategist
+1 212 761-4706 +1 212 761-0222
Kyle.Yellin@morganstanley.com Catherine.Liu2@morganstanley.com

2
M
North America Insight

Contents
6 Executive Summary – The Growing Role of Credit Markets in Financing GenAI

15 Sizing the Spend: A ~$2.9tr Opportunity

19 Technology Credit: Expect ~$200bn of Bond Issuance to Complement Sizable Cash Flows and New Leases

27 Securitized Credit: An Increasingly Sought-After Take-out Option

31 An $800bn Opportunity in Private Credit, Led by Asset-Based Finance (ABF)

Morgan Stanley Research 3

M
North America Insight

Bridging a $1.5tr Data Center Financing Gap

Key Takeaways

■ We project a $800bn+ growth opportunity around data center investment via private credit, led by asset-based
finance. This supports the structural growth story in private credit, especially in higher-quality segments.
■ We also view DC securitized credit as a secular growth story that is relatively insulated from potential macro
pressures, and project $150bn of DC ABS/CMBS issuance through 2028.
■ We expect $200bn of incremental IG tech bond issuance, and think credit spreads can remain tight given the
manageable magnitude and higher credit quality.
■ We think credit markets have significant capacity to meet these financing needs with broad AUM growth across
segments. Risks are much slower growth and lower yields.
■ Our equity analysts flag alt. asset managers, banks, and other companies that benefit directly from capital
deployment and growth in private credit (not just limited to AI).

Filling an estimated $1.5tr players may be disinclined to debt-fund capex early in the investment
medium-term financing gap for cycle, leaning more on internal cash flows and leases to start. Faster
data center investments: We revenue monetization (vs. expectations) could mean corporate bond
forecast $2.9tr of capex for issuance is higher than our estimates.
global data centers (DC)
through 2028 across build and Expect securitized credit to play an increasingly important role in
hardware costs. Total invest- financing data centers; forecast $150bn new issuance across DC
ment needs significantly exceed ABS/CMBS through 2028: To date, roughly 10% of US capacity has
our expectation for $1.4tr of hyperscaler capex funded by cash flows backed ~$50bn in issuance in these products, but we expect to see
(leveraging company guidance and Morgan Stanley equity research an increasingly high rate of securitization, with 25% of new GWs in
estimates). In effect, we see a $1.5tr financing gap that needs to be the US securitized market by 2028. Indeed, DC developers that have
met by external capital, and expect credit markets to play a big role already begun issuing into securitized credit markets have ambitious
here. growth plans, with their global capacity set to increase substantially
in coming years; they have not yet securitized a majority of their oper-
$800bn opportunity for private credit, led by asset-based finance ational capacity. We see considerable potential for supply in these
(ABF): Private credit sits at the intersection of significant AUM markets, aided by the relatively competitive costs of financing
growth in a higher rates environment on the one hand, vs. complex, achieved through securitization and by leveraging their contracts
large-scale, global, and customized financing needs on the other. with high-IG tenants.
With corporate bond and securitized credit funding limited by issuer
willingness and market capacity, respectively, we estimate private Financing capacity in credit, and risks to our estimates: We think
credit in many forms, but led by ABF will need to fill over half of the credit markets should see continued inflows from multiple channels,
gap identified above. We project a ~$800bn growth opportunity for and should have significant capacity to absorb investment needs; we
private credit through 2028. do not expect a financing bottleneck. Further, we emphasize that pri-
vate credit mandates are growing in higher-quality mandates like
We forecast ~$200bn incremental investment-grade technology ABF, which is well tailored for DC financing needs. The risk in the
bond supply through 2028: Hyperscalers have been regular issuers medium term is a sharp slowdown in growth and significant policy
in the corporate bond market, yet remain under-indexed relative to easing (beyond our base case) in which real yields collapse. This
the stock market, and have significant capacity to increase debt would weigh on new demand, especially from long-term end inves-
without pressuring credit ratings. Yet, we project just $200bn of tors like insurance companies/pension funds.
incremental corporate debt issuance through 2028. We think these

4
M
North America Insight

Exhibit 1: We present five main financing paths for medium-term global data center spend

Source: Morgan Stanley Research

Morgan Stanley Research 5

M
North America Insight

Executive Summary – The Growing Role of Credit

Markets in Financing GenAI
Forecasting ~$3tr global data center spend Beyond the scope of this note, we expect that building out the incre-
through YE28 mental power infrastructure could cost at least $300bn.

Generative artificial intelligence (AI) and tech diffusion remains a key On an annual basis, financing needs increase to just over $900bn in
theme for Morgan Stanley's Research division, shaping market per- 2028. A detailed breakdown of these capex numbers is in the fol-
formance and the economy over the coming years. However, this lowing section . These are large numbers any way we look at them,
transformative technology comes with sizeable investment needs. through both macro or micro lenses. For context, all companies in
We forecast roughly $2.9tr on global data center (DC) spend, not the S&P 500 combined spent ~$950bn on capex in 2024, compa-
inclusive of power investments, through 2028 (see Exhibit 2 ), of rable to our DC-only number for 2028. Our US economists esti-
which roughly $2.5tr will be specifically for AI-related workloads. mate that AI-related capex will contribute 0.4-0.5% to GDP growth
The $2.9tr includes roughly $1.6tr on hardware (chips/servers) and over 2025-26, while our China economists have flagged notable sup-
$1.3tr on building out data center infrastructure, including real port to GDP through 2027.
estate, build costs, and maintenance, among others (see Exhibit 3 ).

Exhibit 2: Total global annual spend on AI and non-AI workloads Exhibit 3: Total global annual spend on DC infrastructure and
Est. Annual Total Global Spend ($mm) chips & servers
$1,000,000
Total GenAI Workload Spend Total non-AI Workload Spend
$900,000
$127,558 Est. Annual Total Global Spend ($mm)
$800,000 $1,000,000
$110,915 Total Capital Cost of Chips + Servers ($m) Total Capital Cost of DCs, Exl. Chips + Servers ($m)
$700,000 $900,000

$600,000 $98,798 $800,000

$500,000 $700,000 $483,827

$87,522
$400,000 $792,706 $600,000 $453,337
$701,789
$300,000 $500,000
$76,740 $564,735 $376,771
$200,000 $405,768 $400,000
$280,081
$100,000 $230,310 $300,000
$0 $200,000 $177,671 $436,437
$359,367
2024e 2025e 2026e 2027e 2028e $286,763
$100,000 $213,210
$129,378
$0
2024e 2025e 2026e 2027e 2028e
Source: Morgan Stanley Research estimates

Source: Morgan Stanley Research estimates

6
M
North America Insight

Estimating incremental financing needs and of investment needs (e.g., long horizon, complex, fragmented,
paths early stage), and a sticky end-investor base (e.g., insurance, HNW
retail, sovereign wealth funds) that needs high-quality asset
The AI-capex cycle has been in motion for some time now, with hyper- exposure well suited to meet these capex needs.
scaler spend alone going from ~ $125bn two years ago to ~$200bn
in 2024, and expected by consensus/guidance to exceed $300bn in In Exhibit 4 , we show our estimate of how the ~$2.9tr capex could
2025. So far, internal cash flows from hyperscalers have been more be broken down across different markets, with each category dis-
than sufficient to match these requirements. Our equity analysts are cussed in detail in later sections.
optimistic about AI monetization, and project that GenAI revenues
could exceed $1tr by 2028, with close to 70% variable margins, com- We caveat that making projections several years out comes with con-
pared to just $45bn in 2024. siderable assumptions and admittedly some guesswork. Clearly, the
premise that underlies our assumptions can shift dramatically over
That said, over the horizon of 2025 through 2028, investment the horizon. Certain forms of financing like sovereign spend can be
needs ramp sharply, hyperscaler cash flow usage is constrained especially hard to estimate. In addition, over a long horizon like we
both by willingness and capital allocation, and there is a lag are predicting, financing can shift from one form to the other (e.g.,
between the timeline of spending vs. monetization. This implies a from asset-based financing to asset-backed securities) depending on
significant gap in capital needs that will need to be financed. several factors, which might mean "undercounting" certain forms of
capital deployment.
In this collaborative report, we look at the role that different capital/
lending markets will play in meeting these financing needs over the These caveats aside, we think that our collaborative approach, which
coming years. In particular, we focus on credit markets – corporate, embeds our equity analysts' assumptions around revenue, capex, and
securitized, private/asset-based lending – which we see gaining trac- cash flows, expertise across different credit markets (top-down and
tion as a more efficient provider of capital. We think there is a favor- bottom-up), and empirical observations provides a reasonable
able alignment of significant and growing dry powder across framework to scope out the numbers.
credit markets with attractive real yields on offer, the ideal nature

Exhibit 4: Our estimated breakdown of data center financing needs (ex Power)

Estimated Financing Breakdown for Total Capex CY25-CY28

Cash Need $2.9tr Global Capex on Data Centers (ex. Power)

$200bn $150bn $350bn Other

Financing $1.4tr Hyperscaler Capex $800bn Private Credit
Corp. ABS & Capital (PE, VC,
Path Funded with Cash Flows Debt CMBS
Opportunity Sovereign)

0 500 1,000 1,500 2,000 2,500 3,000

Estimated Cash Need or Source through 2028 ($bn)

Source: Morgan Stanley Research

Morgan Stanley Research 7

M
North America Insight

Estimate $1.4tr hyperscaler capex may be holder returns, potential acquisitions and allowing for cash build for
funded with cash flows, leaving a $1.5tr optionality may imply that practical estimates for AI-capex spend are
financing gap lower.

Hyperscaler AI capex has been largely self-funded by cash flows Considering our estimate that there could be $200bn of related cor-
so far. Rising cash flows may be able to continue to support elevated, porate debt issuance, as we introduce next, we use $1.4tr in our anal-
and even growing, capex. For this part of our analysis, we focus on ysis to represent capex spend in our broader framework that may be
large-cap technology companies that generate sizable cash flow covered by hyperscaler cash flows. Remaining cash flows may be
from operations that cover their AI- and/or data center-related capex used for shareholder returns and for cash build. Capital allocation
needs. changes, specifically increased shareholder returns and/or acquisi-
tions, can impact this analysis on a single-name idiosyncratic basis.
Capital intensity remains elevated, with the hyperscalers' combined
2025 capex guide at >$300bn. And looking at the four-year period of Overall, this leaves a remaining sizable $1.5tr gap relative to the
2025-2028, our equity research counterparts forecast capex growth ~$2.9tr spend estimate we laid out above. The rest of this spend
and almost $1.6tr of cumulative spend among those players. Over will need to come from external sources of capital. We discuss some
that time frame, our equity research counterparts do forecast cash thoughts and numbers around how this $1.5tr gap could land across
flow from operations to grow and sum to almost $3tr. Thus, while different markets next.
there is capacity to cover more spend via cash flows, we think share-

Exhibit 5: Large-cap corporate cash flows are significant, but Exhibit 6: GenAI monetization could have implications for timing
after accounting for shareholder returns and potential cash build, of outside funding
still leave a gap to our broader 4Y capex estimates Hyperscaler Revenue & EBITDA Expectations CY25-CY28 Estimated by
Morgan Stanley Equity Research
Hyperscalers' Cash Flows & Uses CY25-CY28 Estimated by Morgan 2,500
Stanley Equity Research Revenue
EBITDA
2,000

Cash Flow from Ops ~$2.9tr Cash Flows from Operating Activities

1,500
$bn

1,000

~$700bn FCF
~$1.6tr Hyperscaler Capex ~$550bn
After
Primary Cash Uses (Use $1.4tr as Proxy / Plug for Shareholder 500
Shareholder
Broader Analysis) Returns Returns

0
2023 2024 2025E 2026E 2027E 2028E
0 500 1,000 1,500 2,000 2,500 3,000
Est. Cash Flow or Use ($bn) Source: Company filings, Bloomberg, Morgan Stanley Equity Research (select models by Keith Weiss and
Brian Nowak). Note: Not AI and/or data center specific.
Source: Company filings, Bloomberg, Morgan Stanley Equity Research (select models by Keith Weiss and
Brian Nowak). Note: Not AI and/or data center specific.

8
M
North America Insight

$200bn of corporate debt issuance – Therefore, we think that ~$200bn is a reasonable estimate to use
willingness to issue debt a bigger constraint in our financing spend calculation to represent potential tech-
than ability nology corporate debt issuance. We find support for this number
specifically from a technical lens, both from the micro and macro
On the surface, corporate debt provides a scalable and cheap form of perspectives. First, we see sizable capacity for major tech hyper-
financing, with yields around 5%. The US investment grade corporate scaler issuers to grow their index debt levels toward the balances of
bond market is well established and nearly $9tr in size. The tech- the largest non-financial IG issuers (i.e., telecom/cable players AT&T,
nology sector has grown as a percentage of that market, from just 2% CMCSA, VZ), with ~$150bn of capacity across just three issuers.
two decades ago to 9.6% now, and hyperscalers have been estab- Second, the credit market is significantly under-indexed to the space,
lished issuers for a decade or more. as the technology index contributes <10% of the IG index’s market
value (hyperscalers themselves ~2%), while the IT sector contributes
We do think that the US IG unsecured public market is set up well ~30% of the market cap of the S&P 500 (hyperscalers ~20%).
to take on any necessary funding for AI capex needs out of tech- Therefore, the IG Technology Index has room to grow its contribution
nology names. The technology sector is relatively higher credit and could reasonably support an additional $200bn of net supply
quality than other IG sectors, illustrated by its lower net leverage and through YE28, pushing its contribution to 12% of the broader IG
tighter trading levels, thereby implying capacity to add debt, in our market all else equal. The IG Technology Index even saw $120bn of
view. We see >$500bn of illustrative issuance capacity across just amount outstanding growth over the past four years.
three large-cap technology issuers, using simple math against their
respective downgrade adj. debt/EBITDA thresholds for current rat- Given current cash flow coverage of capex for many players and
ings. This could even be understated given expectations for con- the still relatively nascent role of tech in the credit world, we
tinued EBITDA growth. expect reliance on unsecured corporate debt to grow in phases
and accelerate later in our horizon through 2028. In this early stage
All that said, actual issuance has been quite low over the past few of the investment cycle, we think hyperscalers will lean on cash
years, despite rising capex needs (see Exhibit 8 ). In our view, this flows, leases, and bespoke financing solutions with alternative asset
comes down to the nature and horizon of investment needs and the managers. But as monetization accelerates, we think investors will
risks around them. We think that large-cap technology companies have a clearer line of sight to facilitate unsecured debt issuance. All
may be averse to significant on-balance sheet debt funding for capex, this said, we think management teams will also consider cost of
which could pressure price/valuation multiples. In particular, hyper- funding and issuance windows. The average cost of issuing IG debt,
scalers might take a measured approach around significant debt- even at today's elevated yields, is ~5%, much lower than comparable
funding for data center build (as opposed to hardware), given long- yield on leasing data centers from developers (which is typically high
term risks. single digits-to-double digits). We expect management teams will be
opportunistic around issuing in times of lower yields, just like we saw
in the '19-'21 period.

Exhibit 7: Hyperscalers account for less than 2% of the IG Exhibit 8: Hyperscale $ bond issuance has been fairly low in
market, roughly 20% of the equity market recent years

Hyperscalers as % of Total Market ($bn) Hyperscaler $ Bond Issuance

20%
Hyperscaler Issuance
18%
Equity % of S&P 500
16% 35 Hyperscaler Issuance as % of Total IG Issuance (RS) 10%
Debt % of IG Market 30
14% 8%
12% 25
10% 20 6%
8% 15 4%
6% 10
2%
4% 5
2% 0 0%
2013

2014

2015

2016

2017

2018

2019

2020

2021

2022

2023

2024

2025

0%
2013 2015 2017 2019 2021 2023 2025
Source: Morgan Stanley Research, Bloomberg Source: Morgan Stanley Research, Bloomberg

Morgan Stanley Research 9

M
North America Insight

It makes sense for us to focus on IG large-cap tech companies and 300MWs in other countries – roughly 10% of the US data center
their possible USD unsecured bond issuance in this report as a capacity. As the markets mature, demand grows, and liquidity
starting point. However, we do see numerous other financing paths improves, we see up to 25% of the MWs in data centers backed by
for players and also issuers outside of these large-cap tech names. ABS or CMBS by 2028. Indeed, many of the issuers which have
Specifically, we acknowledge leases as another form of financing, already used the ABS or CMBS markets have ambitious growth plans.
recent IG tech issuance in other currencies, recent IG tech issuance We estimate that around 45% of the data centers that are opera-
for AI-related M&A, data center REIT unsecured issuance ahead, and tional and owned by ABS issuers currently back outstanding issu-
more. ances; of the data centers that are planned, operational, or under
construction, we estimate around a quarter back ABS deals (by
Read more later in this report: Technology Credit: Expect ~$200bn count). Moreover, these issuers alone have likely securitized to date
of Bond Issuance to Complement Sizable Cash Flows and New just a tenth of their planned global capacity. Indeed, 14 of 17 ABS
Leases . issuers for which data were available combine to have nearly 25GW
of planned total capacity; a higher reliance on the ABS or CMBS mar-
Securitized credit markets – data center ABS kets could see these markets grow considerably. Thus, we don't
and CMBS market to finance ~$150bn view supply as a constraint on growth, and indeed we think we will
through 2028 see issuers increasingly tap the ABS/CMBS markets once they
have sufficiently tested the waters.
Post-construction and tenant occupancy, we think securitized credit
markets will play an increasingly important role in supplying capital Issuing into the securitized credit market provides the issuer with
at reasonable levels to data center developers. While these devel- a more competitive cost of capital for a comparable duration than
opers will likely raise initial financing from a handful of other sources it might have been able to achieve elsewhere through private
(e.g., bank lending, construction loans), we expect many developers credit, construction financing, or indeed unsecured credit. Since
to leverage ABS or CMBS markets as a takeout option once tenants 2022, issuers in the ABS market have had an average financing cost
begin to pay rent, securitizing the properties themselves or the cash of around 5.5%, which we expect would be lower than they would
flows from the leases. Given the deeper and more established secu- achieve should they have issued into the unsecured market. Indeed,
ritization markets in the US, we expect that the opportunity from the US ABS coupons are comparable to USD IG debt (4-7y) coupons
securitized credit will be greater for US-based assets in the early and are significantly lower than the average HY coupon over the
stages of the investment cycle. However, we expect to see expan- same time period. At present, we think these issuers would be
sion of transactions to include non-US based assets over time. unlikely to achieve an IG rating; the securitized credit markets thus
Already, Canadian data centers have been included in some US allow for them to directly leverage their contractual relationships
deals, and there have been two European deals to date. with IG-rated tenants and issue debt at lower coupons.

At present, we have seen approximately $50bn of securitized credit Read more later in this report: Securitized Credit: An Increasingly
for data centers with around ⅔ in ABS and ⅓ in CMBS. We estimate Sought-After Take-out Option .
that this represents around 4GW of power in the US, with a further

Exhibit 10: Securitized credit market financing rates are closer to

Exhibit 9: Estimated annual issuance of data center-backed
IG funding costs than HY
securitized credit could reach >$60bn p.a. by 2028 ($ in bn)
US DC ABS US HY Corp 4-7y
Estimated Annual Issuance for Data Center Securitized Credit ($bn) 10.00%
$70 US IG Corp 4-7y, A/BBB US DC CMBS Fixed Rate
9.00%
ABS Issuance CMBS Issuance
$60 8.00%
$50 $25 7.00%
6.00%
$40
$19 5.00%
$30 4.00%
$20 $13 3.00%
$37
$7 $25 2.00%
$10
$14
$10 1.00%
$0
0.00%
2025 2026 2027 2028
2018 2019 2020 2021 2022 2023 2024 2025

Source: Morgan Stanley Research estimates

Source: Morgan Stanley Research, Intex

10
M
North America Insight

An $800bn opportunity for private credit, Exhibit 11: The total addressable market (TAM) for ABF is signifi-
led by ABF (asset-based financing) cantly larger than direct lending

As discussed above, our corporate credit and ABS market forecasts

add up to roughly $350bn of financing needs through 2028, which
even when added to the $1.4tr of hyperscaler cash flows, leaves a
sizeable gap to the total $2.9tr number. We think private capital –
in particular, credit – will play a key role in meeting a majority of
the remaining financing gap as it sits optimally at the intersection of
significant expansion in AUM in a higher rate environment and the
complex, global, and customized financing needs that are associated
with AI build-out.

Taking a step back, it is important to clarify what we mean by private

credit. While this term is often conflated with corporate loans to
middle market companies, the universe of private credit is much
larger. It also encompasses private IG placements, asset-based
finance (ABF), infrastructure lending, mezzanine loans, construction
loans, among others. Private credit has been a significant growth area
for institutional asset managers over the past few years (see Exhibit
12 ). But importantly, much of the end asset growth is coming from Source: Morgan Stanley Research estimates. Note: *Direct lending fund AUM from Cobalt; current
deployed capital for ABF does not include securitizations.
investors like insurance companies, sovereign wealth funds and pen-
sion funds that need high-quality asset exposure (as opposed to
lending to highly-levered companies).

Exhibit 12: Private credit AUM has grown sharply in recent years

Global Private Debt Fund AUM

$2,000 Total AUM ($B) % Deployed (rs) 80%
$1,800 70%
$1,600
60%
$1,400
$1,200 50%
$1,000 40%
$800 30%
$600
20%
$400
$200 10%
$0 0%

Source: Morgan Stanley Research, Pitchbook LCD, Bloomberg. *Note: Data as of 6/30/2024

Morgan Stanley Research 11

M
North America Insight

The case for further expansion in private credit AUM remains strong agreements between remaining public data center REITs (typically
across institutional and retail channels, especially if rates stay ele- BBB-rated) and private equity capital—most notably, a $15 billion+ JV
vated vs. their post-GFC averages. Insurance cos. AUM is growing rap- between Equinix (EQIX) and GIC/CPP.
idly, and retail investors are significantly "under-exposed" to private
assets. With an increasing appetite for returns that are more stable Meanwhile, AI remains a top priority for venture capital. According to
and uncorrelated to macro drivers, we think there is a clear case for PitchBook, nearly 60% of global VC investment in 1Q 2025—totaling
further expansion. $73 billion—was directed toward AI and machine learning startups.
This share was even higher in North America, with more than half of
What form could private credit financing take around AI infra- the deal value concentrated in just three companies: OpenAI,
structure? We think asset-based finance will play a key role here. Anthropic, and Groq.
We discuss more details in the private credit section, but for readers
who are not familiar, ABF is a broad term for financing agreements Another source of capital that could be sizeable in building data
that are secured by loans and contractual cash flows such as leases center financing needs is sovereign spend. We already have seen
with or without hard assets, a concept widely used in securitizations. several announcements to this effect, from Saudi Arabia, France,
Unlike construction loans specific to a single entity or a single South Korea, China, among others.
project, ABF is a broader category in which loans are secured by a
broader pool of assets that generate predictable cash flows. Ultimately, how these investments are funded (i.e., direct outlays vs.
sovereign fund investments vs. debt being raised in markets) remains
We think this style of lending is tailored to meet data center financing unclear. If it happens through the channel of sovereign wealth funds/
needs. At a very basic level, it provides a beneficial arrangement for public pension funds, we think that effectively is a form of capital
many parties involved. Hyperscalers/other companies get to lease a market financing, which gets captured by the categories we discussed
data center without incremental cash usage or debt funding. The above. On the other hand, if governments are willing to spend surplus
investor earns significant current income (HSD to low double digit), cash or raise sovereign debt for this spend, then that is an additional
as the cash flows on the leases are coming from highly credit-worthy form of financing to consider.
companies (in most cases). And for the developer, it allows a more
efficient form of financing than using highly-dilutive equity capital. Ultimately, there are too many moving parts/unknowns in these
"other" categories, including bank lending, to have micro precision.
Our equity analysts have flagged alt. asset managers and banks Instead, what we assume is that the $1.15tr gap left after excluding
that can benefit directly from capital deployment and growth in pri- hyperscaler cash flows/bonds/securitized credit is split 70/30
vate credit (not just limited to AI). Key beneficiaries are discussed between private credit (discussed above), and the other sources. We
here. think the 70/30 split has some intuition based on 1) debt-equity mix
of typical project financing/construction loans, and 2) empirical evi-
Read more later in this section: An $800bn Opportunity in Private dence based on debt financing within observed strategic partnerships
Credit, Led by Asset-Based Finance (ABF) . between corporates and alt. asset managers.

Other capital sources – $350bn across We expand on this in more detail later in the An $800bn Opportunity
sovereigns, PE, VC, bank lending, and others in Private Credit, Led by Asset-Based Finance (ABF) section.

Similar to credit markets, there is substantial dry powder and Risks, other considerations, and implications
active fund-raising underway for equity capital. While notable
AI-related equity investments are emerging, they are primarily 1) Significant capacity in the credit market; we don't expect a
concentrated in data center joint ventures between REITs and pri- financing bottleneck
vate equity firms, as well as in venture capital financing.
Publicly-disclosed examples of private equity participation have In the securitized credit section later, we discuss the possibility of
largely taken the form of data center REIT take-privates and subse- power bottleneck in data center build-out as a real risk over time.
quent JV partnerships. Over the past five years, following transac- Given the scope of this report, it also worth discussing that could
tions such as KKR/GIP’s acquisition of CONE, we have observed JV there also be a financing bottleneck considering how large some of

12
M
North America Insight

the numbers are. For instance, the $800bn private credit number we Exhibit 13: Insurance companies are seeing significant inflows
have estimated compares to the entire HY bond market at $1.4tr in this cycle with higher rates, supporting demand for credit
size. ($bn) ($bn)
75 Fixed Rate Inflows and Outflows 50
60 40
At the risk of stating the obvious, much depends on how the
45 30
macro environment shapes up, but based on our assessment of 30 20
credit conditions today, we don’t see a lack of financing capacity. 15 10
At a very high level, credit markets are seeing inflows across several 0 0

investor types, life insurance companies, pension funds, and overseas (15) (10)
(30) (20)
buyers. In contrast, credit issuance has been fairly tepid over the past Inflows (LHS) Outflows (LHS) Net flows (RHS)
(45) (30)
three years. The high level of rates, along with rolling uncertainties

3Q16
1Q17
3Q17
1Q18
3Q18
1Q19
3Q19
1Q20
3Q20
1Q21
3Q21
1Q22
3Q22
1Q23
3Q23
1Q24
3Q24
(e.g., inflation, geopolitics, trade policy), have meant that issuance
Source: LIMRA, Morgan Stanley Research
has been crowded out from credit markets. If at all, credit markets
have a "too much demand, not enough supply" problem right now.
Exhibit 14: Insurance companies are large end-owners of credit
More specifically, within overall credit, private credit is seeing signifi- markets
cant demand. And as flagged above, higher-quality segments are US Corporate Bond Market Ownership
Households/Other Insurance Mutual Funds Pension Funds Financial Instit. Others Non-US Investors
growing, rather than purely non-IG business loans. In this context,
100.0%
the right comparison for our $800bn private credit number is not
80.0%
the leveraged finance market, but instead the investment grade
60.0%
market (which is $9tr).
40.0%

20.0%
The total asset-based financing (ABF) market is around $25tr, and 0.0%
90 92 94 96 98 00 02 04 06 08 10 12 14 16 18 20 22
with this, the "private" portion (i.e., by non-banks) is already at $2tr.
Source: Federal Reserve, Morgan Stanley Research
As alt. asset managers continue to focus on ABF, we think there will
be significant growth in this market. In fact, as we noted in a
BluePaper last fall, ABF has become the new frontier for insurance rates could also hurt demand from pension funds and overseas inves-
balance sheets, allowing private credit managers to originate a tors as the value proposition in USD-denominated credit erodes. So
broader set of assets and place them through structuring and credit this is certainly a risk worth watching, especially as our economists
enhancement that make these assets suitable for investment. expect the Fed to cut 7 times next year.

2) What could challenge the capacity of credit markets to provide But we want to clarify that the rate decline would have to be quite
this financing? substantial and well beyond our economists' base case expectation.
While this topic requires an in-depth report by itself, we flag that
The macro environment matters here, and slower growth would nat- despite lower rates, our forecasts see real rates remaining positive
urally raise some fundamental concerns. However, the related bigger and yield-curve remaining steep. We think risks of sharply weaker
risk, in our view, is if rates fall sharply. In particular, the growth in cap- inflows rises if/when rates drop significantly below 3%.
ital inflows into credit has been supported by steep yield-curves, and
positive real rates. These have driven significant growth into insur- 3) Time to monetization can affect funding choices
ance products, including but not limited to fixed-rate annuities. In
addition, while corporate bonds remain the biggest allocations for Our estimates across the different financing opportunities are
insurance companies, they are increasing the share of securitized broadly based on: 1) headline capex needs, 2) hyperscaler cash flow
credit and private credit notably. and revenue estimates through 2028, 3) capital allocation decisions
around debt-funded capex, and 4) strategic trends such as rapid
If rates go down sharply, we think this channel of significant inflows growth in IG-rated private credit mandates. Should revenue growth
could be disrupted. As shown above, fixed annuity inflows were quite be higher and faster than our equity analyst estimates, financing
negligible in the pre-COVID era when yields were much lower. Lower paths could shift.

Morgan Stanley Research 13

M
North America Insight

In particular, we think the public credit market could play a bigger Exhibit 15: Capex growth was sizeable through the late 1990s, and
role in this scenario as companies become more comfortable with it, debt funding was a big driver
as it remains the cheapest source of financing. On the other hand, ($bn)
700
Historical IG Gross Issuance
weak macro conditions and slower revenue growth could mean even
600
more reliance on external capital, specifically private credit.
500

400
4) Long-term macro implications of significant capex + credit
growth — some similarities to the 1990s, but many differences 300

200

The last time a capex cycle of this magnitude played out was back in 100

the mid-to-late 1990s, driven by investment in hardware, technology, 0

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
and web services. Seasoned credit investors will recall that through
the later part of the decade, corporates were very aggressive with Source: Dealogic, Morgan Stanley Research

their capital allocation, and debt/leverage increased sharply. The

period ultimately ended with the sharpest default cycle in recent his-
tory from 2000-04. Exhibit 16: The cycle culminated with a meaningful default cycle

14.0%
All Corporates Default Rate Speculative-grade Default Rate
Arguably, it is too early in the current investment cycle to be con-
12.0%
cerned about risks on the other side. Clearly, debt growth of this mag-
10.0%
nitude run into the risk of obsolescence, slower monetization, and
8.0%
weak macro conditions. An encouraging sign, though, is the diverse
6.0%
pools of capital that are available today, which can distribute the
4.0%
warehousing of credit risk (unlike the late 1990s when it was con- 2.0%
centrated on corporate balance sheets). Further, the ultra-high- 0.0%
1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

2010

2012

2014

2016

2018

2020

2022
quality credit profile of hyperscalers and significant cash on hand
mean less sensitivity to macro conditions.
Source: Moody's, Morgan Stanley Research

14
M
North America Insight

Sizing the Spend: A ~$2.9tr Opportunity

Sizing investment spend on AI

We expect that data center growth in the US and globally will accelerate rapidly in the coming years, driven predominantly by the greater
need for high-performance compute (HPC). Morgan Stanley Research estimates that the total power used by GenAI will rise from 29TWh
in 2023 to 686TWh in 2028 globally. This expansion of power for GenAI will drive total data center power usage (including traditional work-
loads like cloud/enterprise) from 370TWh in 2023 to 1235TWh in 2028 (or said another way, from 56GW of capacity in 2023, up to 188GW
in 2028). We expect that about 55% of the GenAI power will be in the US. For more details on our approach for sizing the power demand from
GenAI, please see our reports here and here. Please ask us for a copy of our GenAI model and our financing model.

Exhibit 17: Total power used by GenAI (TWh), Global Exhibit 18: Total power used by GenAI (TWh), US
800TWh 200%
Total Power Used by GenAI (TWh) Annual Growth Rate 400TWh Total Power Used by GenAI (TWh) Annual Growth Rate 500%
700TWh 180%
350TWh 450%
160%
600TWh 400%
140% 300TWh
350%
500TWh 120% 250TWh 300%
400TWh 100%
200TWh 250%
300TWh 80%
150TWh 200%
60%
200TWh 150%
40% 100TWh
100%
100TWh 20% 50TWh 50%
0TWh 0%
0TWh 0%
2023e 2024e 2025e 2026e 2027e 2028e
2023e 2024e 2025e 2026e 2027e 2028e

Source: Morgan Stanley Research estimates Source: Morgan Stanley Research estimates

We forecast that the cost of building new data centers, for both AI and non-AI workloads, could total ~$2.9tr cumulatively from 2025-
2028 on a global basis, or ~$1.7tr in the US alone. To size the market, we break down the spend into two distinct categories: the cost of the
chips and servers, and the cost of everything else ( Exhibit 20 ). We build our model by using our power estimates and by estimating the dollar
cost per MW for both the build and the chips that will be hooked up to the power; as such, the costs are placed in the year the data center
becomes operational, but we note of course that data center constructions are multi-year projects, and some of this spend will be spent ahead
of the calendar year of operation.

Exhibit 19: Our global capex estimates, split out by type of spend (build vs hardware), region, and workload type.
Global Capex on Data Centers ($bn, 2025-2028)

Total Capex 2025-2028

$2,890

Global Build Costs Total Hardware Costs

$1,296 $1,594

US APAC Europe ROW AI Hardware non-AI Hardware

$754 $101 $117 $324 $1,419 $175
~$11mm/MW ~$8mm/MW ~$12mm/MW ~$10mm/mW ~$16.5mm/MW ~$6.5mm/MW
Source: Morgan Stanley Research estimates

Morgan Stanley Research 15

M
North America Insight

Total spend on hardware will come to around $1.6tr, by our esti- Total spend on the build will be around $1.3tr, by our estimates.
mates. We expect that spend on Nvidia (and other) GPUs and custom For the non-hardware spend, we expect that costs to construct a new
silicon (+servers) will average around $17mn/MW over the next four data center will vary by region, with some of the lowest global costs
years, assuming 8 chips per server. The mix of the types of chips sold in Southeast Asia. Included in those costs are the costs of cooling, the
will change over time, as will the power consumption per chip, and land, building fit-out and mechanical costs, the networking equip-
we expect costs to fall marginally on a $/MW basis through 2028 ment, transceivers, cabling, and other costs. We also expect that the
(from $19.2mm/MW in 2024 to $13.7mm/MW in 2028). We build our costs of a data center suitable for the AI workloads will be more
estimates using quantity, power, and cost estimates for over 35 types expensive than non-AI workloads.
of AI chips. We expect spend on non-AI chips and servers will be
around $7mn/MW. Note that we expect that between 60-80% of the
growth in power will be for GenAI workloads, with the balance for the
non-AI workloads.

Exhibit 20: We estimate a spend of $2.9tr for new data center construction between 2025 and 2027, globally
Global 2024e 2025e 2026e 2027e 2028e
Gen-AI Workload
Capital Cost of DCs, Excluding Chips + Servers ($m) $83,331 $160,887 $228,227 $294,489 $362,345
Capital Cost of GPUs/ASICs + Servers ($m) $146,978 $244,881 $336,509 $407,300 $430,361
Total GenAI Workload Spend $230,310 $405,768 $564,735 $701,789 $792,706
Cumulative Spend AI Spend 2025-2028 $405,768 $970,503 $1,672,292 $2,464,998

Non-AI Workload
Capital Cost of DCs, Excluding Chips + Servers ($m) $46,047 $52,323 $58,536 $64,878 $74,092
Capital Cost of non-AI Chips + Servers ($m) $30,693 $35,199 $40,262 $46,037 $53,466
Total non-AI Workload Spend $76,740 $87,522 $98,798 $110,915 $127,558
Cumulative Spend non-AI Spend 2025-2028 $87,522 $186,321 $297,236 $424,794

2024e 2025e 2026e 2027e 2028e

Total Capital Cost of DCs, Exl. Chips + Servers ($m) $129,378 $213,210 $286,763 $359,367 $436,437
Total Capital Cost of Chips + Servers ($m) $177,671 $280,081 $376,771 $453,337 $483,827
Total ($m) $307,049 $493,290 $663,534 $812,704 $920,264
Cumulative Spend (2025-28) ($m) $493,290 $1,156,824 $1,969,528 $2,889,792
Source: Morgan Stanley Research estimates

Exhibit 21: Total global annual spend on DC infrastructure and Exhibit 22: Total global annual spend on AI and non-AI workloads
chips & servers Est. Annual Total Global Spend ($mm)
$1,000,000
Total GenAI Workload Spend Total non-AI Workload Spend
$900,000
Est. Annual Total Global Spend ($mm) $127,558
$1,000,000 $800,000
Total Capital Cost of Chips + Servers ($m) Total Capital Cost of DCs, Exl. Chips + Servers ($m) $110,915
$900,000 $700,000
$800,000 $600,000 $98,798
$700,000 $483,827 $500,000
$87,522
$600,000 $453,337 $400,000 $792,706
$701,789
$500,000 $300,000
$376,771 $76,740 $564,735
$400,000 $200,000 $405,768
$280,081
$300,000 $100,000 $230,310

$200,000 $177,671 $436,437 $0

$359,367
$286,763 2024e 2025e 2026e 2027e 2028e
$100,000 $213,210
$129,378
$0
2024e 2025e 2026e 2027e 2028e
Source: Morgan Stanley Research estimates

Source: Morgan Stanley Research estimates

We expect that 55% of the GenAI workload will be situated in the US, with a total estimated spend of ~$290-500bn (including non-AI
DCs). In the US, we assume that the data center costs, excluding the GPUs and servers, will be around $12mn/MW for GenAI and slightly lower
at $10mn/MW for non-AI DCs. We see the main risk to our view that 55% of the data center workloads will be in the US is the power bottleneck
(described more below); we would expect that should there be insufficient capacity in the US to power data centers, that capacity would likely
be sourced elsewhere in the world, and that best efforts would be made to get chips online and operational somewhere.

16
M
North America Insight

Exhibit 23: In the US, we expect a cumulative spend of ~$1.6tr between 2025 and 2028
US $mm/MW 2023e 2024e 2025e 2026e 2027e 2028e

Total Data Center Power Growth (MW) 7,189 11,205 14,678 18,154 21,194
GenAI Data Center Power Growth (MW) 4,334 8,065 11,223 14,355 17,014
Non-AI DC Power Growth (MW) 2,855 3,140 3,454 3,800 4,179

Gen-AI Workload
Data Center Cost, Excl. GPUs + Servers $12
Capital Cost of DCs, Excluding Chips + Servers ($m) $52,012 $96,784 $134,681 $172,257 $204,173
Capital Cost of GPUs/ASICs + Servers ($m) $83,398 $138,036 $189,171 $228,743 $232,501
Total GenAI Workload Spend $135,410 $234,819 $323,852 $401,000 $436,675
Total Cumulative GenAI Spend , 2025-2028 $234,819 $558,671 $959,671 $1,396,346

Non-AI Workload
Data Center Cost, Excl. GPUs + Servers $10
Capital Cost of DCs, Excluding Chips + Servers ($m) $28,546 $31,401 $34,541 $37,995 $41,795
Capital Cost of non-AI Chips + Servers ($m) $17,416 $19,841 $22,634 $25,855 $28,885
Total Non-AI Workload Spend $45,962 $51,242 $57,175 $63,850 $70,679
Total Cumulative non-AI Spend , 2025-2028 $51,242 $108,417 $172,267 $242,947
2024e 2025e 2026e 2027e 2028e
Total Capital Cost of DCs, Exl. Chips + Servers $80,558 $128,185 $169,222 $210,252 $245,968
Total Capital Cost of Chips + Servers $100,813 $157,877 $211,805 $254,598 $261,386
Annual Total $181,372 $286,062 $381,027 $464,850 $507,354
Cumulative Spend, 2025-2028 $286,062 $667,089 $1,131,939 $1,639,293
Source: Morgan Stanley Research estimates

Exhibit 24: Total annual US spend on DC infrastructure and chips Exhibit 25: Total annual US spend on AI and non-AI workloads
& servers Est. Annual Total US Spend ($mm)
$600,000
Total GenAI Workload Spend Total Non-AI Workload Spend
Est. Annual Total US Spend ($mm)
$600,000 $500,000
Total Capital Cost of Chips + Servers Total Capital Cost of DCs, Exl. Chips + Servers
$500,000 $400,000

$400,000 $300,000

$200,000
$300,000

$100,000
$200,000

$0
$100,000
2025e 2026e 2027e 2028e

$0
2025e 2026e 2027e 2028e
Source: Morgan Stanley Research estimates

Source: Morgan Stanley Research estimates

We expect most of the AI data center growth to be new builds, with an estimated cost of around $21-$31mn/MW (including GPUs +
servers) or ~$8-$14mn/MW (excluding GPUs + servers). However, the time to power for a new build can be multiple years from the time
the shovel first hits the ground, with power challenges particularly acute in the US. A report from the Lawrence Berkeley National Lab high-
lighted that "projects built in 2022 took 5 years from the interconnection request to commercial operations, compared to 2 years in 2015 and
<2 years in 2008." We expect that the opportunity to convert existing large load assets, particularly those that are currently used for Bitcoin
mining or the conversion of old Industrial facilities, will be an important part of the build in the near term as data center developers look for
solutions that can provide a time advantage for bringing their hardware online faster. However, even if a powered shell is purchased from
another use, significant costs will need to be spent on the retrofit; we estimate that the cost of the powered shell, plus the electrical, building
fit-out, and mechanical work accounts for ~15% of the total data center cost (including hardware).

Morgan Stanley Research 17

M
North America Insight

Risks to growth: underappreciated and overstated risks

Access to power in the US is the number one risk to growth at the Exhibit 26: Potential shortfall of power for US data centers, 2025-
moment. We estimate that there is a potential shortfall of ~45GW 2028
of power in the US through 2028; a recent survey by Schneider Potential Shortfall in Power for US Data Centers, 2025-28
70
Electric found that 92% of respondents see grid constraints as a chal-
lenge, with long utility wait times slowing down expansion. (The 60

Schneider Electric survey respondents further cited permitting 50

issues, fiber availability, and access to chips as the next three causes
40
of slowed down data center construction.) This power bottleneck is
30
driven by a number of factors, including long lead times to secure
new grid interconnection, power equipment shortages, an aging 20

power transmission system, and greater weather extremes placing 10

strains on grid reliability, among other reasons.
-
US DCs Under Construction Potential Shortfall
US Power Needed, 2025-28 Available US Grid Capacity
There are a number of de-bottlenecking solutions that will likely
Source: Morgan Stanley Research estimates
need to be drawn upon to fill this 45GW gap, including 1) converting
crypto mining sites into data centers (which could provide >10GW of
suitable US capacity; however, this option becomes more expensive Moreover, while we have heard concerns that AI infrastructure
with a rising cost of bitcoin), 2) siting data centers at large US nuclear capex will not yield an attractive ROI for the key players, thus
power plants (20-25 GW of suitable US capacity, though with signifi- slowing down/reducing development, our equity research col-
cant regulatory uncertainty for "behind the meter" data centers, leagues do not share that view. They expect that GenAI yields a posi-
which could reduce this solution down to 10-15GW), 3) building new tive ROI starting this year and with total benefits over $1tr by 2028
natural gas-fired power plants (though capacity here is constrained (see their BluePaper here). More specifically, they expect that GenAI
by long build times and waitlists), and 4) on-site fuel cells (see more could drive a $1.1tr revenue opportunity in 2028 (up from $45bn in
details here). 2024) with margins approaching 70%. In addition, they estimate the
consumer impact of GenAI on e-commerce, search, social, autono-
We think concerns that hardware and software efficiency gains mous, and wearables will drive an additional $680 billion of spending
will reduce the overall demand for compute (and hence data cen- by 2028. With this monetization of GenAI on the horizon, we expect
ters) are overblown. While some unconfirmed media reports sug- that data center demand is likely to remain very strong, and the
gest that some of the major players (here and here) are pulling back spend is justified by the fundamentals.
from lease discussions, and investors may interpret that news as a
harbinger of a wider problem of data center overcapacity, we do not
think this is the case, and we expect any excess capacity to be picked
up by other players. Further, developments with respect to step-
function improvements in the cost of training large language models
should usher in a more rapid transition from training to inference
rather than stymie the growth in data center development. Indeed,
our equity research colleagues' channel checks suggest that demand
for GPUs remains strong, and there are signs of continued strength
in US AI infrastructure spend from the major hyperscalers.

18
M
North America Insight

Technology Credit: Expect ~$200bn of Bond Issuance to

Complement Sizable Cash Flows and New Leases
Large-cap tech cash flows may cover ~$1.4tr of our 4Y AI spend estimate

As the starting point for our entire financing path breakdown, we represents potential corporate debt issuance. Therefore, we think it
estimate $1.4tr of spend through YE28 may be funded with cash is fair to use a $1.4tr number in our analysis to represent capex
flows of large-cap tech companies. To date, we've seen hyperscaler spend in our broader framework that may be covered by hyper-
AI capex largely self-funded by cash flows. We expect that could con- scaler cash flows.
tinue near term.
We recognize that our analysis is basic as not all capex is necessarily
To help size funding needs, the hyperscalers' combined 2025 capex for data center and/or AI spend, and we are excluding adjacent
guide is >$300bn. This 2025 run-rate could imply >$1.2tr of cumula- spending. Overall, this leaves a remaining sizable $1.5tr gap relative
tive capex spend for 2025-2028. Our equity research counterparts to the ~$2.9tr spend estimate.
forecast almost $1.6tr of cumulative spend during that time frame,
with embedded YoY growth ( Exhibit 27 ). Sizable spend is specifi- It is worth noting as well that current cash balances in this space
cally for technology infrastructure, like servers, data centers, and are already sizable. These provide an additional cushion for spend
networking. aside from future cash flow generation. For example, just Alphabet
(GOOGL) and Microsoft (MSFT) combined have $175bn of cash, cash
Theoretically, the multi-year sum of cash flow from operations is equivalents, and short-term investments as of 3/31/25.
high enough to cover the total capex spend we laid out earlier. But a
combination of capital allocation decision-making (e.g., investments, Shareholder returns and acquisitions are other key uses of cash.
shareholder returns, acquisitions) and lack of visibility into long-term For many of the large-cap technology companies, shareholder
funding needs may mean that, in practice, the amount of cash flow returns are sizable but manageable. However, capital allocation
that may be earmarked for capex may be much lower, with excess changes, specifically increased shareholder returns and/or acquisi-
cash after other financial policy priorities going to build the cash bal- tions, can impact this analysis on a single-name idiosyncratic basis as
ance. changes could increase the need for debt funding and/or lessen the
focus on investment. In our view, the use of cash flow for investment,
Hence, we lean into our equity research capex estimates, specifically rather than shareholder returns, is constructive from the credit per-
that $1.6tr Morgan Stanley equity research 4Y aggregated capex spective, something we flagged for ORCL specifically in our recent
number, to come up with the internal cash flow component of our note.
financing breakdown. From that number, we remove $200bn, which

Exhibit 27: Cash flow from ops should continue to cover capex needs for hyperscalers in the medium term, per Morgan
Stanley equity research forecasts
1,000 Cash flow from ops Capex Shareholder returns FCF after shareholder returns
900
Cash flow estimated sum by MS ER across

800

700
hyperscalers ($bn)

600

500

400

300

200

100

0
CY23 CFO CY23 Cash CY24 CFO CY24 Cash CY25E CFO CY25E Cash CY26E CFO CY26E Cash CY27E CFO CY27E Cash CY28E CFO CY28E Cash
Use Use Use Use Use Use

Source: Company filings, Bloomberg, Morgan Stanley Equity Research (select models by Keith Weiss and Brian Nowak)

Morgan Stanley Research 19

M
North America Insight

Technology sector credit quality & ratings provide significant issuance cushion, on paper

From a fundamental lens, the technology sector has stronger credit quality than other sectors across the IG market, thereby implying
capacity to add debt, in our view. The sector carries considerably lower leverage ( Exhibit 28 ), with 0.8x net leverage vs. the broader non-finan-
cial IG space up at 1.8x and other sectors in the 1.5x-2.0x (excl. outlier Utilities). We also take note of a significantly higher cash/debt ratio, a
higher average credit rating (A2/A3), and other supportive metrics. We see this credit quality reflected in valuations, with the tech index trading
at a ~15bp premium to the IG index ( Exhibit 29 ).

Exhibit 28: The technology sector has notably lower leverage than Exhibit 29: Technology remains one of the tightest trading sec-
other IG sectors tors, trading at a ~15bp YTD average premium to broader IG,

Select Credit Metrics by Sector at 1Q25 reflecting its credit quality

Gross Net Cash /

Sector 350 IG Index (LUAC)
Leverage Leverage Debt IG Tech Index (I00394US)
Communication Services 2.4x 2.0x 40% 300
Consumer 2.4x 1.9x 18% 250

Index OAS (bps)

Energy 1.8x 1.5x 11%
200
Health Care 2.4x 1.7x 27%
Information Technology 1.8x 0.8x 56% 150
Industrials 2.1x 1.7x 16%
100
Materials 2.3x 1.8x 15%
Utilities 5.6x 5.6x 1% 50
Non-Fin IG 2.4x 1.8x 17% 0
Source: Bloomberg, Morgan Stanley Research. See more in the Credit Strategy 1Q25 Fundamentals deck Dec-19 Dec-20 Dec-21 Dec-22 Dec-23 Dec-24
(slide 15).
Source: Bloomberg, Morgan Stanley Research. Updated 7/10/25, using a 5-day ma. Note: IG Technology
Index excludes select hyperscalers but is a good proxy.

We see >$500bn of illustrative issuance capacity across just three issuers, we arrive at a sum of >$500bn of illustrative issuance
issuers (and >$200bn for just one), using simple math against capacity. For Alphabet specifically, at the time of its USD/EUR bond
their respective downgrade thresholds for current ratings. For deals this spring, S&P called out that the company would have to
each corporate issuer, each rating agency provides not only a rating increase its net debt by >$200bn before reaching the 1x adj. net
for the name, but also specific commentary around how an upgrade leverage downgrade threshold for the AA+ rating (4/28/25 note).
or a downgrade may be achieved. Typically that commentary focuses
on debt/EBITDA ratios, but can also include FCF/debt and/or qualita- In this simple analysis, we may be understating capacity in two ways.
tive factors. In this case, we focus on three issuers that had the most First, we do account for near-term EBITDA growth through FY25, but
sizable FY24 capex spend. These are very highly-rated names, and we're not looking through YE28. Second, we assess capacity just
S&P provides a similar 1-1.5x sustained net adj. debt/EBITDA metric based on the current ratings level. It is possible companies may be
as a threshold across the respective downgrade commentaries. We willing to fall to a lower ratings tier. We saw lower-quality ORCL let
found that the downgrade threshold at Moody's for these is more itself fall a few tiers a few years back as it debt-funded shareholder
qualitative, hence our focus more on S&P. returns and an acquisition. However, on the other hand, it is impor-
tant to clarify that the "debt" we quantify here is not just unsecured
In our illustrative adj. debt cushion analysis in Exhibit 30 , we calcu- bond issuance, but would include increases in leases (debt-like liabili-
late an estimate for additional adjusted debt cushion, using FY25E ties included in agency-adjusted debt balances), as we discuss later .
EBITDA and accounting for current adjusted debt. For just three

20
M
North America Insight

Exhibit 30: We present >$500bn of illustrative issuance capacity across just three issuers, using a simplistic approach against adj. debt/
EBITDA downgrade thresholds for current ratings
Illustrating Sizable Debt Cushion Under Select Tech Bond Ratings' Downgrade Thresholds
Issuer GOOGL MSFT AMZN
Agency for analysis S&P S&P S&P
Rating AA+ AAA AA
Sustained lev >1x, as result Lev >1x sustained, most Lev exceeds 1.5x, among
Select downgrade threshold details (focus more on
of acq. or change in likely from sustained heavy other metrics &
quantitative rather than qualitative)
financial policy capex & large acq. considerations
Reported recent adj. leverage NM 0.2x 0.4x
Reported recent adj. debt ($bn) (54.8) 29.2 58.6
Agency est. FY25E adj. EBITDA ($bn) 172.2 161.2 155.9
Illustrative issuance capacity ($bn) 227.0 132.0 175.2
FY25 end for reference 12/31/25 6/30/25 12/31/25
Select info above as of… 12/31/24 6/30/24 12/31/24
Source: Note dated… 3/6/25 5/21/25 5/6/25
Source: Company filings, S&P, Morgan Stanley Research.

Companies may lack willingness for significant & immediate on-balance-sheet funding but will
be opportunistic

We think technology companies would like to avoid major debt All this said, in regards to time frame, we think management
funding, such as the hundreds of billions implied by ratings teams will still be opportunistic around cost of funding and issu-
cushion capacity, and we would expect reliance on unsecured cor- ance windows, even if issuance is not extreme in size. In the tech
porate debt to grow in phases. We think that large-cap technology space in the past, many management teams have been efficient with
companies may have an aversion to significant on-balance sheet debt a quantitative approach to capital allocation, issuing debt when rates
funding for capex, which could pressure price/valuation multiples. are lower (and most often using that to buy back stock and/or build
This has implications not only for the size of balance sheet debt, but cash). This is illustrated by the three-year-period between late 2018
also for the timeline for such funding too. and late 2021, when the IG Tech Index YTW fell significantly by
almost 2 percentage points and index debt jumped by more than 50%
We think equity investors would probably prefer a piecemeal and lim- ( Exhibit 31 ). Currently, the IG index is sporting a 5.1% YTW, which
ited approach vs. sizable jumps in unsecured on-balance-sheet debt, compares to recent lows of 4.6% in September 2024 and five-year
especially when ROI is in question and there are risks of tech obsoles- highs of 6.4% in fall 2023, but is still well above the tights of the past
cence. Given current cash flow coverage of capex for many players, decade in the 2% context ( Exhibit 32 ). Our macro strategists see
and the still relatively nascent role of tech in the credit world, we interest rates moving notably lower, forecasting 10Y UST yields at
expect reliance on unsecured corporate debt to grow in phases, and 3.45% 1Y out and the 30Y at 4.15%, while our US Credit Strategists
accelerate later in our horizon through 2028. In this early stage of the forecast little movement in IG spreads with their base case of 90bp
investment cycle, we think hyperscalers will lean on cash flows, cash at 2Q26 end.
on hand, leases, and more. But as monetization accelerates, we think
investors may have a clearer line of sight to facilitate unsecured debt
issuance.

Morgan Stanley Research 21

M
North America Insight

Exhibit 31: We saw IG Tech index debt jump, as index YTW Exhibit 32: The IG tech index is still yielding ~1.4 percentage
dropped, in the '18-'21 time frame, but still grew by $120bn in the points more than the past-decade-average
'21-'25 time frame 7.5 IG Tech BBB Index (I00348US)
900 10 7.0 IG Tech Index (I00394US)
IG Tech Index Amt Outstanding (LHS)
6.5
800 IG Tech Index Yield to Worst (RHS) 9 IG Tech Index Last 10Y Avg
6.0
8 5.5
700
5.0
Amt Outstanding ($bn)

Index YTW (%)

7 4.5
600
6 4.0

YTW (%)
500 3.5
5 3.0
400 2.5
4
300 2.0
3 1.5
200 2 1.0
0.5
100 1 0.0
0 0 Jul-15 Jul-17 Jul-19 Jul-21 Jul-23 Jul-25
Jul-05 Jul-09 Jul-13 Jul-17 Jul-21 Jul-25 Source: Bloomberg, Morgan Stanley Research. Updated 7/10/25, using a 7-day ma.

Source: Bloomberg, Morgan Stanley Research. Updated 7/10/25, using a 3-day ma.

Single-name index levels and tech sector index weighting provide reasonable framework for ~
$200bn issuance capacity plug

We think that ~$200bn is a reasonable estimate to use in our between a potential $85bn cap and the current index debt balances
financing spend calculation to represent potential technology corpo- of three issuers, those with the most sizable CY24 capex spend
rate debt issuance. This is lower than what we may see from the fun- ( Exhibit 33 ). We arrive at ~$150bn cushion summed across just
damental perspective above, i.e. the >$500bn of cushion vs. ratings those three.
thresholds. But we think our view on technology issuer willingness
to fund with on-balance sheet debt is an important qualitative con- This may be a conservative approach given 1) numerous other tech-
sideration. And further, we find support for a $200bn estimate for nology issuers, 2) a potential argument for a cap >$85bn supported
IG tech corporate issuance from a technical lens in two ways: 1) by the higher credit quality of many tech issuers relative to lower-
from a micro perspective, assessing single-name index debt levels rated Comcast (A3/A-/A-) and AT&T (Baa2/BBB/BBB+), and 3) higher
relative to telco/cable issuers, and 2) from a macro perspective, telco index debt balances in the past, with AT&T's par value contribu-
using the IG Technology Index weighting and historical amount tion to the index exceeding >$100bn back in 2017. But this approach
outstanding growth. does seem more reasonable than our earlier ratings-implied capacity
math.
First, from a micro perspective, we see capacity for major tech
issuers to grow their index debt levels toward telco/cable bal- This idea of index debt levels has come up in the context of ORCL
ances, specifically ~$150bn of capacity for just three of them. credit, a larger and lower-quality balance sheet that did a US$7.75bn
When thinking about unsecured debt issuance, we think it is reason- six-part bond deal earlier this year (though proceeds were largely for
able to cap technology issuers' index debt balances at the largest refinancing). We discussed in a recent note how the total debt
non-financial corporate issuers' index debt balances, considering growth we expect over the next two years, if all eligible, could put
when the market might be "full" from a technical standpoint. Those ORCL in contention for the top nonfinancial issuer index spot vs.
largest names happen to be in the TMT space, but in Communications telco peers, AT&T specifically. Read more: TMT Credit Research:
(AT&T and Comcast), with both posting total par value index debt of ORCL Credit: Opening the (Star)Gate to More Capex, Bond Issuance,
~$80bn+. Our illustrative debt cushion math assesses the difference and Leases (17 Jun 2025).

22
M
North America Insight

Exhibit 33: We arrive at ~$150bn of illustrative issuance capacity across just three technology credits when we solely focus on index debt
of each vs. the top non-financial IG index constituents; in this exhibit, we list the top IG index constituents in order of par value index debt
outstanding

US Corporate IG Index (LUAC) - Top 25 Members

Par Index Par Index
Ticker GICS Sector Ticker GICS Sector
Debt ($bn) Debt ($bn)
JPM Financials 177.7 DUK Utilities 66.8
BAC Financials 175.1 TMUS Communication Services 62.0
MS Financials 144.8 AVGO Information Technology 56.9
C Financials 118.7 ABBV Health Care 56.4
WFC Financials 117.9 CHTR Communication Services 52.8
GS Financials 114.8 AMGN Health Care 52.2
T Communication Services 86.8 GM Consumer Discretionary 50.7
HSBC Financials 84.0 AMZN Consumer Discretionary 50.3
CMCSA Communication Services 80.8 CVS Health Care 50.1
ORCL Information Technology 80.0 NEE Utilities 49.9
VZ Communication Services 78.7 PFE Health Care 49.6
AAPL Information Technology 75.5 F Consumer Discretionary 46.3
UNH Health Care 73.4

Illustrative Index Debt Cushion for Select Tech Credits vs. Top Non-Financial IG Index Constituents
Issuer GOOGL MSFT AMZN
Imposed index debt cap (approx T/CMCSA balance) 85.0 85.0 85.0
Current index debt of hyperscaler 16.0 42.6 50.3
Implied index debt technical cushion 69.0 42.4 34.8
Source: Bloomberg, Morgan Stanley Research. Notes: Bold tickers in top table denote names covered by MS TMT Credit Research (all by Lindsay Tyler, with exception of CHTR coverage led by David Hamburger). AVGO
balance includes VMW. Updated 7/10/25.
++ Rating and estimates for this company have been removed from consideration in this report because, under applicable law and/or Morgan Stanley policy, Morgan Stanley may be precluded from issuing such informa-
tion with respect to this company at this time.

Further, from a broader macro perspective, the IG Technology We also think it is supportive that the sector has seemingly held back
Index has room to grow and could reasonably support an addi- on borrowing for some time but is slowly starting to ramp back up.
tional $200bn of net supply through YE28, in our view. The sector The growth of tech debt outstanding, relative to the broader IG
has now grown over the past two decades, from ~2% to 9.6% of the market, has slowed since 2020, but year-to-date, has moved 0.4 per-
US IG corporate credit market now, illustrating the growing impor- centage points higher. Further, the technology sector's role in credit
tance of the sector to investors and impact on the market broadly still pales in comparison vs. its ~30% market cap contribution to the
( Exhibit 34 ). This current contribution to the IG market roughly S&P 500. Lastly, the index grew by ~$120bn in amount outstanding
aligns with that of healthcare/pharma, communications, and utilities. during the last four years (shown in Exhibit 31 ).

We think the ratio of IG Tech Index/IG Index debt could increase, even All in all, while select data center players sit outside of the IG Tech
beyond the high of ~10.1% in spring 2023. In our simplified illustrative Index, we think it is a good proxy to confirm our initial micro analysis
sensitivity analysis in Exhibit 35 , we find that the ratio could move technical takeaways.
up to ~12% if ~$200bn of net new IG technology index-eligible debt
was hypothetically added to the IG index. This is a conservative
approach assuming no other net supply in other sectors, so such a
bump could yield a lower ratio.

Morgan Stanley Research 23

M
North America Insight

Exhibit 34: Tech debt has grown to represent almost 10% of Exhibit 35: Sensitivity analysis to illustrate the impact of $0-200bn
the broader IG market, while ~30% of S&P 500 market cap is in of additional IG tech index debt on its weighting in the IG index
information technology names
Impact of Add'l Tech Debt, All
Index Balances For Reference
40% Info. Tech. Index as % of S&P500 (Mkt Cap) Else Equal
Tech Index as % of US IG Index (Mkt Value) Additional Tech
35% Implied IG Tech Current Mkt
Index Debt Index Name
30% / IG Index Ratio Value ($bn)
($bn)
25% 0 9.6% IG Tech
678
50 10.3% (I00394US)
20%
100 10.9% IG Index
7,049
15% 150 11.5% (LUACTRUU)
10% 200 12.1%
Source: Bloomberg, Morgan Stanley Research. Updated 7/10/25.
5%

0%
Jul-00 Jul-05 Jul-10 Jul-15 Jul-20 Jul-25
Source: Bloomberg (using I00394US, LUACTRUU, S5INFT, SPX), Morgan Stanley Research. Updated
7/10/25.

Notable lease liability uptick by select IG hyperscalers, a debt-like liability, in our view

Data center leases are another funding vehicle for corporates' MSFT's lease liabilities between operating and finance have more
AI-related spend, and we're seeing MSFT and ORCL use these in than doubled since CY22, and it has "not yet commenced" lease bal-
size. As we discussed in our recent note on ORCL credit, for the ance peaked above $115bn in mid-2024 ( Exhibit 36 , Exhibit 37 ).
hyperscaler names, leases are typically for facilities such as data cen- Meanwhile, we've seen ORCL's operating leases more than double
ters and offices. A few benefits of using leases are that companies are over the past two years, while additional lease commitments that
looking to address additional data center capacity needs quickly haven't yet commenced stand at $43bn.
(build vs. rent), and this part of the financing doesn't get rolled up
into balance sheet debt. ORCL management discussed on its last Rating agencies treat leases as an adjustment added for the
earnings call that it has building partners that charge the company adjusted debt balance. Therefore, the additional lease commit-
rent once data center construction is finished. But when ORCL has ments may result in a more sizable differential between reported bal-
suddenly higher capex, it means the company is filling out data cen- ance sheet debt and rating agency-adjusted debt balances.
ters, buying components to build computers, and putting them on
the floor.

Exhibit 36: MSFT's lease liabilities, across operating and Exhibit 37: MSFT's additional finance leases that haven't yet
financing, have more than doubled since CY22 end and now commenced, primarily for DCs, peaked above $100bn in
rival its reported balance sheet debt number CY2Q24 and CY3Q24
70 Total operating lease liabilities 140 Additional finance leases - not yet commenced
MSFT Disclosed Add'l Leases Not Yet
MSFT Bal. Sheet Leases, Q End ($bn)

Total finance lease liabilities Additional operating leases - not yet commenced
60 120
Commenced, Q End ($bn)

50 100

40 80

30 60

20 40

10 20

0 0

Source: Company filings, Morgan Stanley Research Source: Company filings, Morgan Stanley Research
24
M
North America Insight

Other AI-related corporate issuance by DC REITs & HY neocloud, or for AI-related M&A, or in
other currencies

This report is meant as a high-level starting point on this topic, so it High-yield bond & loan issuance: For example, neocloud company
makes sense for us to be focusing on IG hyperscalers and their pos- CoreWeave (CRWV) priced a $2bn 5NC2 at par to yield 9.25% in late
sible USD unsecured bond issuance. However, we do see adjacent May 2025, following its private credit deals and IPO. Use of proceeds
financing paths, cash uses, and issuers. Our comments below were earmarked for general corporate purposes, including repay-
regarding EQIX's funding plans, hyperscaler issuance in other cur- ment of outstanding indebtedness.
rencies, and new HY neocloud issuance serve to substantiate our
~$200bn corporate issuance plug. IG tech bond issuance for AI-related M&A: While not directly for
data center spend, we have seen IG technology issuance over the past
Data center REIT unsecured issuance: Over the past few years, few years for acquisitions where AI was mentioned in the deal
we've seen data center REIT take-privates (ex. KKR/GIP for CyrusOne announcement releases. For example, Cisco (CSCO) did a $13.5bn
in a $15bn deal), and those developers are now issuers in the ABS seven-part bond deal in February 2024, with use of proceeds for gen-
market rather than in the corporate market. The current eral corporate purposes, including to partially finance the acquisition
USD-denominated bond balances of the two remaining large of Splunk. Hewlett Packard Enterprise (HPE) did a $9bn six-part bond
IG-rated data center operators, Equinix (EQIX) pales in comparison deal in Sept. 2024, with use of proceeds to help fund the recently-
relative to the hyperscalers, with $8.9bn of index debt relative to closed Juniper acquisition. Synopsys (SNPS) did a $10bn six-part
MSFT's $42.6bn. EQIX has a JV in place with GIC and CPP, and has bond deal in March 2025 with net proceeds to fund part of the cash
been active in issuance in other currencies, given the lower cost of portion of the still-pending Ansys deal.
funding. However, we could see more issuance in USD in the future.
Specifically, in March 2025 at the MS TMT Conference, EQIX man- Presence of preferreds and converts as alternative: For example, in
agement outlined its plans to be in the debt capital markets in the US connection with the aforementioned acquisition of Juniper, HPE
and highlighted 1-2 turns of rating agency capacity for leverage. EQIX priced $1.35bn of Series C Mandatory Convertible Preferred Stock in
then hosted an Analyst Day in late June 2025 and disclosed plans to September 2024. Super Micro Computer (SMCI), a maker of AI
raise $16bn of debt over the next five years, including 2025, with servers that doesn't have corporate unsecured bonds outstanding,
$8bn of refinancing and $8bn of incremental debt. was recently in the market with a new zero-coupon convertible due
in 2030.
IG tech bond issuance in other currencies: For example, Alphabet
(GOOGL) did a US$5bn four-part bond deal and a €6.75bn five-part
bond deal in April 2025.

Morgan Stanley Research 25

M
North America Insight

Worth following hyperscalers' involvement in venture capital, major AI funding partnerships,

power partnerships

Paired with their direct investments in data centers, Big Tech compa- Partnerships with asset managers, funds, other corporates: We
nies are getting involved in venture capital funding, especially of have seen not only JVs between data center REITs and private equity,
large headline start-ups, partnerships/projects with other corpo- but also large fund announcements between hyperscalers and other
rates and funds, and power agreements with players like in the players. The AI Infrastructure Partnership (AIP) was formed by
nuclear space. These developments are worth following and can act BlackRock, Global Infrastructure Partners (GIP), Microsoft, and MGX
in the form of current/future cash uses or sources. last year. Further, the Stargate project was announced in January
2025.
IG tech companies' involvement in venture capital: More broadly,
AI investments remain a priority for venture capital, with PitchBook IG tech companies' agreements with power players for data cen-
flagging almost 60% of global VC dollars invested in 1Q25 going to ters: Our colleagues earlier highlighted the potential shortfall of
AI and machine learning startups (with a higher percentage for North power in the US in the medium term. There is rising demand for elec-
America). As reported by Business Insider, of the >$90bn in deal tricity by data centers, and we've seen Big Tech make requests for
value in the US in 1Q25, more than half was accounted for by OpenAI's power and announced agreements with power players. For example,
$40bn round, Anthropic's two funding rounds totaling $4.5bn, and Talen Energy announced in June 2025 an expanded nuclear energy
Groq's $1.5bn round; large-cap corporates have been involved in partnership to supply electricity from a Pennsylvania plant to
funding for all three (ex. MSFT in OpenAI since 2019 and AMZN in Amazon AWS data centers. Another example was the announcement
Anthropic). Corporates continue to be significant contributors of between Microsoft and Constellation Energy in the fall.
venture dollars, alongside VC firms, with PitchBook highlighting that,
of the $192bn in venture dollars invested in the US in 2024, corporate
venture capital (CVC) participated in $108bn worth of deals.

26
M
North America Insight

Securitized Credit: An Increasingly Sought-After Take-out

Option
We estimate that the securitized credit data center market – including here ABS and CMBS – could provide around $150bn of financing
between 2025-2028 to the US data center industry. Data center leases, on a per megawatt basis, are the underlying source of income for
the data center owners. These owners are likely to achieve a more favorable cost of financing by leveraging these contractual relationships
with investment grade tenants.

Securitized credit products are generally issued only once data centers have achieved some level of stabilized performance state and are
operational with tenants in place. The financing raised by these issuances are often used to refinance the developers' existing debt, including
project finance or construction loans necessary to build the data center in the first place, or to fund the development of new construction
projects. More recently, however, we have seen a CMBS issuer (GSMS 2025-800D) experimenting with securitizing financing backed by a data
center that is under construction, the first securitized credit data center deal to feature a data center that is highly transitional in nature.

Sizing the securitized credit opportunity

Exhibit 38: Supply of data center securitized credit bonds in 2025 has already surpassed previous
annual highs

Gross New Issuance of Data Center Securitized Credit ($mm)

$14,000
CMBS ABS $12,150
$12,000 $11,699
$10,355
$10,000 $9,555

$8,000

$6,000

$4,000 $3,219
$1,733 $1,861
$1,700
$2,000

$-
2018 2019 2020 2021 2022 2023 2024 2025 YTD

Source: Finsight,Trepp, Morgan Stanley Research. Issuance data as of June 30, 2025.

We estimate that over $50bn of data center-backed debt has been The securitized credit market for data centers will supply around
raised in USD securitized credit to date since 2018, with roughly $150bn in ABS and CMBS issuance from 2025-2028, by our esti-
⅔ ($36bn) through the ABS market and a further ⅓ ($17bn) in the mate. Using disclosures from the deal documents, we estimate that
CMBS markets.1 There have also been two European ABS deals for roughly 2.6GW underpin the US ABS deals (with a further 300MWs
data centers in the UK and Germany issued in recent years for a total in Canada, UK, and Germany) and around 1.4GW underpin the CMBS
of roughly $1.1bn. Overall, data center backed securitized debt is deals. We estimate that there are around 42GW of installed data
growing rapidly with supply in 2025 already surpassing the entirety center capacity in the US, and so we believe that the securitization
of 2024 (~$12.2bn this year vs $11.7bn last year). We anticipate that market to date backs just under 10% of total capacity in the US. We
the ABS and CMBS markets will be an increasingly attractive option think the rate at which new GWs that come online will be securitized
for data center developers through which they refinance their will increase in the coming years as the markets mature, and hence,
existing debt or fund new developments as the market matures and we expect to see an annual securitization rate up to 25% across both
liquidity improves. ABS and CMBS markets by 2028.

1. Note that figures are as of June 30th, 2025.

Morgan Stanley Research 27
M
North America Insight

Exhibit 39: ABS and CMBS supply forecast through 2028 Exhibit 40: Securitized credit issuance vs newly-installed DC
Estimated Annual Issuance for Data Center Securitized Credit ($bn) capacity
$70
ABS Issuance CMBS Issuance
$60
$70
Total Securitized DC Issuance Est. and Newly Installed US DC Capacity 25
$50 $25 $60
20
$40 $50
$19
$30 $40 15

$20 $13 $30 10

$37
$7 $25 $20
$10
$14 5
$10 $10
$0
2025 2026 2027 2028 $0 0
2025 2026 2027 2028

Source: Morgan Stanley Research estimates Total Securitized Issuance ($bn) US New Installed DC Capacity (GW)

Source: Morgan Stanley Research estimates

Most of the issuers which have already come to market have ambi- Exhibit 41: Est. securitized capacity vs global ambitions, sample
tious growth plans. We estimate that around 45% of the data cen- of leading ABS/CMBS issuers
ters that are operational and owned by ABS issuers currently back MWs in Securitized Deals vs Global Capacity Ambition, Leading Issuers
6000
outstanding issuances (with an average rate per developer of ~50%); Est. Securitized Capacity by ABS
5000
of data centers that are planned, operational, or under construction, 4000 Est. Securitized Capacity by CMBS
we estimate just around a quarter back ABS deals (by count) (with an 3000 Announced Capacity Remaining

average rate per developer of 35%). Moreover, these issuers alone 2000

have likely securitized to date just around 10% of their planned 1000

global capacity (17% on average). Indeed, 14 of 17 ABS issuers for 0

Aligned Stack Vantage Switch CyrusOne
which data were available combine to have nearly 25GW of planned
Source: Deal docs, corporate disclosures, Morgan Stanley Research. Data as of June 30, 2025.
total capacity; a higher reliance on the ABS or CMBS markets could
see these markets grow considerably. Thus, we don't view supply as
a constraint on growth, and indeed we think we will see issuers The securitization market could potentially be a global solution,
increasingly tap the ABS/CMBS markets once they have suffi- though it is highly concentrated in the US for now. The ABS market
ciently tested the waters. thus far has primarily financed US data centers, with ~90% of MWs
underpinning the deals located in the USA and the balance spread
At least three issuers have also used both the ABS and CMBS mar- out across Canada, the UK, and Germany. Many of the US data center
kets, including Switch and CyrusOne. Indeed, we think to date that developers also operate campuses in Canada (primarily British
preference for one market over another has been largely due to Columbia, Ontario, and Quebec); most also have global growth ambi-
familiarity with the market rather than because of a fundamental tions with developments across Europe and APAC. We thus see the
preference for one structure over another. Moreover, we have heard possibility of increased CAD, GBP, or EUR data center ABS supply in
that issuers foresee large financing needs over the coming decade coming years, but we would be hesitant to quantify the expansion as
and thus prefer to test out the waters, so to speak, in multiple mar- we think the limitations there are primarily demand-side. While
kets sooner rather than later. Over time, we think it is likely that ABS recent European Commission's legislative proposals for revamping
markets lend themselves more to developers that lean heavily into securitization regulation, following last year's consultation, make us
diversity (both in terms of type and quantity of tenants, and also optimistic about the revival of European securitization markets, we
number of data center campuses in each deal), while the CMBS temper our enthusiasm pending the finalization of the regulatory
market may be better suited, through its SASB structure, to large changes due next year (see here for more details from our European
training facilities with few or one hyperscaler tenants. That said, securitized analyst, Vasundhara Goel).
there is crossover between the types of collateral found via the two
financing structures, and so we think that the collateral usage may be The securitization market has also only financed the build of data
fluid across both ABS and CMBS going forward. centers rather than the underlying hardware, but we see an
opportunity for securitization of GPUs going forward too. Tenants
primarily lease space (on a per MW basis) and provide their own hard-

28
M
North America Insight

ware (the servers and chips). However, there is an emerging class of achieve should they have issued into the unsecured market. Indeed,
companies that provide GPUs as a service (GPUsaaS), such as the US ABS coupons are comparable to USD IG debt (4-7y) coupons,
CoreWeave. These companies lease out GPUs to hyperscaler clients, and significantly lower than the average HY coupon over the same
and we think these contracts could ultimately be securitized into an time period. At present, we think these issuers would be unlikely to
ABS structure with the hardware leases backing the deals. Unlike achieve an IG rating; the securitized credit markets thus allow for
data center ABS, we would expect a GPU ABS to be amortizing given them to directly leverage their contractual relationships with
the useful shelf life of a chip (4-8y, for example) compared to a data IG-rated tenants and issue debt at lower coupons. As these devel-
center (>20y). opers mature, it is possible that they could transition to preferring
unsecured debt once they would be eligible for an IG rating.
Relative cost of financing
Spreads should tighten in the next 12 months, in our view. We think
Issuing into the securitized credit market provides the issuer with existing pools of demand are robust, particularly from insurance
a more competitive cost of capital for a comparable duration than funds and other yield buyers, and we think there are potential
it might have been able to achieve elsewhere through private pockets of demand from securitized credit investors that are still sit-
credit, construction financing, or indeed unsecured credit. Since ting on the sidelines as they ramp up from an education standpoint
2022, issuers in the ABS market have had an average financing cost on the space.
of around 5.5%, which we expect would be lower than they would

Exhibit 42: Average financing cost over time vs gross issuance, US Exhibit 43: Average CMBS cost of financing on fixed-rate bonds
DC ABS market over time
Total Gross Issuance Average Financing Cost CMBS Issuance Conduit Fixed Rate SASB Fixed Rate
$10,000 6.00% $6,000 8%

$9,000 7%
5.00% $5,000
$8,000
6%

Average Financing Cost

$7,000
Gross Issuance ($bn)

4.00% $4,000
5%
$6,000
$5,000 3.00% $3,000 4%
$4,000 3%
2.00% $2,000
$3,000
2%
$2,000
1.00% $1,000
1%
$1,000
$- 0.00% $0 0%
2018 2019 2020 2021 2022 2023 2024 2025 2018 2019 2020 2021 2022 2023 2024 2025

Source: Finsight, Morgan Stanley Research. Data as of June 30, 2025 Source: Trepp, Morgan Stanley Research Estimates. Data as of June 30, 2025

Exhibit 44: New issue US DC ABS vs IG and HY corporate spreads Exhibit 45: Recent new issue data center ABS vs other ABS prod-
Data Center ABS New Issue Spreads vs IG, HY Corp Index ucts
1200
IG Corp Index
HY Corp Index New Issue Spreads of ABS Products, April-May 2025
1000 10%
AAA
Fiber BB
800 AA
9% Subprime Auto BB
A
600 BBB 8%
BB
Yield (%)

400 Fiber BBB Solar A

7% Consumer & MPL BBB
Fiber A Data Center A Solar AA
200 Prime Auto BBB
6% Data Center AAA
Consumer & MPL AAA Subprime Auto BBB
0 Subprime Auto AAA Aircraft AAA Prime Auto A
Jan-18 Jan-19 Jan-20 Jan-21 Jan-22 Jan-23 Jan-24 Jan-25 5% Prime Auto AAA
Prime Auto AA
4%
Source: Finsight, Bloomberg, Morgan Stanley Research. Data as of June 30, 2025 - 1 2 3 4 5 6 7 8
WAL (Years)

Source: Finsight, Morgan Stanley Research. Data as of June 30, 2025

Morgan Stanley Research 29

M
North America Insight

Exhibit 46: US ABS coupons and CMBS fixed rate vs HY and IG USD corporate cou-
pons over time

US DC ABS US HY Corp 4-7y

10.00%
US IG Corp 4-7y, A/BBB US DC CMBS Fixed Rate
9.00%
8.00%
7.00%
6.00%
5.00%
4.00%
3.00%
2.00%
1.00%
0.00%
2018 2019 2020 2021 2022 2023 2024 2025

Source: Finsight, Bloomberg, Morgan Stanley Research. Data as of June 30, 2025

Risks to market growth of capacity being built will not be needed. Quantum computing, as a
further extreme, could dramatically reduce the demand for tradi-
1. Shift in preference to other debt types: While many data center tional GPUs and re-imagine the entire compute industry. Securitized
developers have looked to the securitized credit markets as their pri- products investors are concerned that the data centers that collater-
mary source of capital from capital markets in recent years, we think alize current ABS structures may see weak demand in five years when
it is possible that a maturation of the industry could result in 1) more their refinancing dates come due, either because of a secular industry
favorable terms in private credit or 2) higher credit ratings, and hence challenge (broad decline in demand for compute) or because of a
a shift to unsecured corporate credit. We believe the data center local challenge (decreased demand for data centers in a particular
players have large ambitions and will require significant financing to region, for example due to more demand for inference compute
meet their goals, but it is possible that they come to view securitized (which needs to be proximal to major urban centers) over training
markets as less preferable to some other debt options; we highlight (which can be situated in remote areas)).
two examples below.
3. Power constraints: As mentioned in this report, the main bottle-
• Aligned (ADC), for example, issued ABS four times between
neck to US data center development at present is the availability of
2021 and 2023 for a total of $2.6b with 323MW across 7 data
power. Chip demand is global, though, and so we believe that if there
centers in its mastertrust structure with a weighted-average
is insufficient power capacity in the US that chips will find power in
cost of 4.2%. Aligned has a stated goal to deploy 5GW+ of
other regions with excess power (even if on a 1-2y delay). However,
capacity, and we estimate that 12 of its 30 announced data
the securitization market is a predominantly American phenomenon,
centers are already operational. However, ADC has not
and our estimates of growth are predicated on our assumption that
returned to public markets since its last ABS transaction in
55% of capacity will be in the US. Should this number prove to be too
2023, and indeed completed a $12bn financing round in
high, we would expect to see lower issuance than we have predicted.
January 2025 ($5bn in equity and $7bn of new debt commit-
ments).
4. Concentration limits: On the demand side, while we believe appe-
• Equinix (EQIX) is one of two of the largest data center devel-
tite for these products is growing and will continue to swell as these
opers in the US, and indeed, there are only two publicly-
products become more mainstream, we are cautious around concen-
traded data center companies. Equinix has not issued
tration limits. Fixed-income investors, we believe, are indeed looking
securitized debt, and instead has issued BBB-rated bonds to
for ways to allocate to GenAI growth, and we think data center securi-
finance its global operations.
tized credit is one important way to do so. However, we think fixed
income funds will have a soft upper limit of tech exposure compa-
2. Data center obsolescence: Some market participants are con- rable to the equity market and will need to gradually build up to
cerned that data center construction will exceed demand for MWs those levels. If pricing becomes less favorable in ABS/CMBS if
over the next decade. Headlines about LLM developer DeepSeek ear- demand wanes, we would expect to see other financing options (such
lier this year prompted concerns that there will be frequent step- as HY unsecured credit, or private credit) take up the workload.
function improvements in LLM training capabilities, and the amount

30
M
North America Insight

An $800bn Opportunity in Private Credit, Led by Asset-

Based Finance (ABF)
With the corporate bond and securitized credit financing numbers above adding to ~$350bn, that still leaves a sizeable gap to the $1.5t number
we raised earlier in the Estimating incremental financing needs and paths sub-section. We think private capital, especially credit, will play a
significant role in meeting this additional gap. Our thesis not based purely on a process of elimination. Instead, we think the type and magnitude
of end-investor demand growth in private credit, evolution of the asset class, and need for high/mid-quality asset exposure make it ideal to
finance data centers early in the investment cycle.

Why private credit is tailored to meet DC financing?

It is helpful to start with a quick assessment of the data center Public data center REITs are another investment channel, but
financing cycle. At the outset, capital is needed for acquiring land to these companies have a very small presence in the public debt and
build the data center facility. Over time, more capital is needed as equity market, after numerous take-privates over the past
construction begins, milestones are met, and the physical infrastruc- decade. For instance, Equinix is a large public data center REIT with
ture/power needs to be set up. Initial financing can often raised in the a market cap of ~$74bn and just $9bn of total debt in the $ IG index.
form of a construction/commercial mortgage loan or project The company is rated within the lower tiers of investment grade
financed by the developer (hyperscaler or another investor) in the across the agencies (mid-to-low BBB).
facility (that provides the equity capital).
Hyperscalers have the capacity and knowledge to build data centers
For smaller-scale projects, bank lending alone may suffice in a form by themselves, but we think there are limits given the size of invest-
of a commercial loan. However, the size of investment needs in aggre- ment needs and risks. According to a recent report from Apollo,
gate are very large, compared to the total outstanding C&I lending pie hyperscalers do self-build, but that is typically 20%-45% of their data
from banks, which is around $3tr. In other words, direct bank lending center capacity needs. The exception is Oracle, which only leases for
does not have the capacity to meet the investment pie. Further, these data centers, although the company does spend on capex for hard-
construction loan facilities typically have shorter horizons, so it is fair ware.
to assume that developers would want to take out any initial bank
lending and transition towards deeper/scalable/stickier sources of
credit.

Morgan Stanley Research 31

M
North America Insight

Asset-based finance: the new frontier for insurance balance sheets

In common parlance and often in media articles, private credit gets conflated simply with direct company loans made to businesses, typically
those that are small/mid-sized and sub-investment grade. Private credit, though, is a broad umbrella term that goes well beyond middle-market
lending and includes ABF (asset-based financing), mezzanine lending, distressed loans, private IG placements (the "4a2" market), and realisti-
cally any other bespoke bilateral financing done by a non-banking institution. In the context of financing data centers, we put special
emphasis on the growth in ABF.

Exhibit 47: What is asset-backed finance?

Source: Morgan Stanley Research

Broadly, we consider financing of assets secured by contractual include a portfolio of multiple assets/facilities in different stages of
cash flows, with or without hard assets to fall under the term ABF. development. Given the need for data centers to be located across
Unlike the standardized data center ABS/CMBS structures discussed geographies and jurisdictions, contractual cash flows have a built-in
earlier (see Securitized Credit: An Increasingly Sought-After diversification potential. Given the nascent stage of the industry,
Take-out Option ), ABF structures can be highly bespoke, enabling there is significant scope for innovation in structuring and collateral
customization to fit the specific needs of investors with varying packages.
levels of risk tolerance, returns, ratings, term, and duration. The stan-
dardization in ABS/CMBS structures, driven mainly by rating agen- The ABF construct works well in conjunction with hyperscalers
cies, often require stabilized cash flows with strict requirements of that (as noted earlier) will have a preference to progressively use
lessee characteristics and residual value guarantees in cases where financial leases to ramp their data center access as the investment
hard assets are part of the collateral package. cycle progresses. In such cases, hyperscalers would lease the facility
from the developer, thereby avoiding additional capital outlay that
In the context of data center financing, ABF opens up a range of would reflect on their balance sheets, either as less cash or more
possible structures and financing options to include assets in dif- debt. This financing arrangement is mutually beneficial for the hyper-
ferent stages of development, ranging from data centers with scaler, developer, and investor. Lease costs are typically mid- to high-
fully stabilized cash flows to others that are in early stages of con- single digits (annual 7-10% of the upfront cost), so they're more
struction and build with just land or with land and power access expensive than unsecured debt costs. Yet, these are quite affordable
established. They could be single asset/facility financing, as well as for cash-rich hyperscalers, especially if this means avoiding large cap-
ital spend.
32
M
North America Insight

For the developer that raises cash through ABF, funding costs are End-investors, on the other hand, are receiving very attractive yields,
likely cheaper than the unsecured bond market. It is hard to know the and exposure is diversified from many other risk-assets and some-
exact comparable bond market number, but most private developers what insulated from macro conditions. Even though it is ultimately
may not meet the requirements to raise debt at IG-like ratings, so the developer responsible for paying the end investors, the cash flow
unsecured BB/B yields might be a better reference point. risk (especially when a DC is pre-leased) is to a higher-credit quality
company, i.e., a hyperscaler. Typical hyperscaler-only leases tend to
be significantly long in duration (10+ years), implying that investors
have significant horizon.

Why the growth in ABF within the private credit universe?

As we have noted in Risks, other considerations, and implications , life insurers, especially US life insurers, have seen a considerable rise in
inflows into their tax-advantaged wealth accumulation products with insurance wrappers such as fixed annuities. This significant portfolio AUM
growth needs high-quality spread products, but with little alpha and dispersion in the public IG market with tight spreads. ABF provides insurers
with exposure to potentially investment-grade-rated assets that have significant excess spread vs. the public market, but also provide uncorre-
lated returns, attractive Sharpe ratios, and potential for diversification to sectors otherwise underexposed.

Exhibit 48: Insurance companies are seeing significant inflows Exhibit 49: ABF now accounts for ~3% of insurance investments,
into fixed annuity products implying significant room for growth
($bn) ($bn)
75 Fixed Rate Inflows and Outflows 50
60 40
45 30
30 20
15 10
0 0
(15) (10)
(30) (20)
Inflows (LHS) Outflows (LHS) Net flows (RHS)
(45) (30)
3Q16
1Q17
3Q17
1Q18
3Q18
1Q19
3Q19
1Q20
3Q20
1Q21
3Q21
1Q22
3Q22
1Q23
3Q23
1Q24
3Q24

Source: LIMRA, Morgan Stanley Research

Source: Moody's, Morgan Stanley Research

The need for such assets is by no means limited to life insurance companies. Pension funds, both public and private, as well as foundations and
endowments have looked to private assets to access uncorrelated returns. Thus far, the majority of such allocations have been to private equity
assets. However, higher interest rates and macro uncertainties have resulted in a meaningfully slower private equity deal environment pushing
investors to expand into new areas, including a variety of asset-based financing structures. In our view, data center ABF is among the most
attractive of such opportunities, providing scalable opportunities with diversification potential.

It is also important to highlight the backdrop of an accelerated transition of lending from banks to non-banks (see Extending Credit: The Evolving
Role of Wholesale Banks in Credit Markets for an expansive discussion). In the context of data center financing needs, this transition represents
a strategic inflection point for both banks and non-banks to develop and adapt their business models to deploy multiple long-dated sources
of capital into an investment opportunity with scale. These shifts could lead to significant changes in the industry and lead to a new breed of
partnerships, joint-ventures, and business models.

An example of AI-related ABF private credit that has attracted investor attention recently is CoreWeave's facilities. CoreWeave special purpose
vehicles (SPVs) entered delayed draw term loan facilities, secured by GPUs and other equipment, in summer 2023 and again in spring 2024.
The funding was led by Blackstone, Magnetar, and others. Bloomberg reported that there are investment grade and non-IG tranches (read more
here NSN SDMZMLT0G1KW).

Morgan Stanley Research 33

M
North America Insight

70/30 split among the residual funding needs gets us to $800bn of private credit financing

Drilling into our estimates for private credit financing of data centers, under construction is in Abilene, Texas, as reported by Bloomberg
to refresh, once we add corporate debt ($200bn) and securitized here. Recent disclosures point to >60% debt funding using construc-
debt ($150bn) to the $1.4tr hyperscaler cash flow number, we are still tion loans. At the end of May, Newmark disclosed the second phase
left with a $1.15tr gap in capital needed. Assuming a 70/30 split of a $15bn JV to fund the 1.2 GW AI data center in Abilene, featuring
between debt vs. "other capital" within this $1.15tr gap, we arrive just over $9bn of construction loans ($2.3bn initial and $7.1bn addi-
at our estimate of ~$800bn of potential private credit growth pri- tional).
marily (but not limited to) via the ABF channel.
Another sizable construction loan of note year-to-date was an
Clearly, we are making a fairly simplistic assumption about a large announced $2bn financing to complete the build-out of Novva Data
number over a fairly long time frame. The intuition behind the 70/30, Centers' flagship 175MW Salt Lake City data center, which followed
though, comes from a few different angles. previous equity commitments from CIM.

First, the current environment generally remains challenging for Third, we found that the initial disclosed plans for both the AI
large equity capital deployment despite significant dry powder. Infrastructure Partnership and Stargate project more broadly help
Older vintages have been hard to exit for sponsors and distributed support the approximate 70/30 assumption we are making for the
capital to LPs remains sub-par relative to history. In the public split. The AI Infrastructure Partnership (AIP), formerly known as the
market, IPO activity has been subdued for the past few years, and Global AI Infrastructure Investment Partnership (GAIIP), was formed
qualitatively, we would expect this to accelerate later in the cycle, as by BlackRock, Global Infrastructure Partners (GIP), Microsoft, and
we approach monetization. Further, the cost of issuing debt should MGX last year, with NVDA and xAI joining the partnership this year.
also get cheaper over time as the Fed cuts rates (MS economists The original release states: “The partnership will initially seek to
expect 7 cuts in 2026). Putting all these factors together, we think it unlock $30 billion of private equity capital over time from investors,
is difficult to see equity capital competing 1:1 with credit capital at this asset owners, corporates, which in turn will mobilize up to $100 bil-
stage of the cycle. lion in total investment potential when including debt financing.”

Second, data center loans or many commercial mortgage loans for Separately, the aforementioned Stargate project was announced in
that matter are often originated with a 70-80% loan-to-value, which January 2025, with plans to begin deploying $100bn “immediately”.
validates our assumption. As our TMT credit analysts highlighted in It is our understanding that the project intends to arrange project
a recent note, ORCL Chairman Larry Ellison had commented back in finance on a data center-by-data center basis, with a limited 10-20%
January at the White House that the first of the Stargate projects equity portion.

34
M
North America Insight

Disclosure Section
Mortgage Backed Securities (MBS) and Collateralized Mortgage Obligations (CMO)
Principal is returned on a monthly basis over the life of the security. Principal prepayment can significantly affect the monthly income stream and the maturity of any type of MBS, including
standard MBS, CMOs and Lottery Bonds. Yields and average lives are estimated based on prepayment assumptions and are subject to change based on actual prepayment of the mortgages
in the underlying pools. The level of predictability of an MBS/CMO's average life, and its market price, depends on the type of MBS/CMO class purchased and interest rate movements. In general,
as interest rates fall, prepayment speeds are likely to increase, thus shortening the MBS/CMO's average life and likely causing its market price to rise. Conversely, as interest rates rise, prepayment
speeds are likely to decrease, thus lengthening average life and likely causing the MBS/CMO's market price to fall. Some MBS/CMOs may have "original issue discount" (OID). OID occurs if the
MBS/CMO’s original issue price is below its stated redemption price at maturity, and results in "imputed interest" that must be reported annually for tax purposes, resulting in a tax liability even
though interest was not received. Investors are urged to consult their tax advisors for more information. Government agency backing applies only to the face value of the CMO and not to any
premium paid.
The information and opinions in Morgan Stanley Research were prepared by Morgan Stanley & Co. LLC, and/or Morgan Stanley C.T.V.M. S.A., and/or Morgan Stanley Mexico, Casa de Bolsa,
S.A. de C.V., and/or Morgan Stanley Canada Limited. As used in this disclosure section, "Morgan Stanley" includes Morgan Stanley & Co. LLC, Morgan Stanley C.T.V.M. S.A., Morgan Stanley Mexico,
Casa de Bolsa, S.A. de C.V., Morgan Stanley Canada Limited and their affiliates as necessary.
For important disclosures, stock price charts and equity rating histories regarding companies that are the subject of this report, please see the Morgan Stanley Research Disclosure Website
at www.morganstanley.com/researchdisclosures, or contact your investment representative or Morgan Stanley Research at 1585 Broadway, (Attention: Research Management), New York, NY,
10036 USA.
For valuation methodology and risks associated with any recommendation, rating or price target referenced in this research report, please contact the Client Support Team as follows: US/Canada
+1 800 303-2495; Hong Kong +852 2848-5999; Latin America +1 718 754-5444 (U.S.); London +44 (0)20-7425-8169; Singapore +65 6834-6860; Sydney +61 (0)2-9770-1505; Tokyo +81
(0)3-6836-9000. Alternatively you may contact your investment representative or Morgan Stanley Research at 1585 Broadway, (Attention: Research Management), New York, NY 10036 USA.

Analyst Certification
The following analysts hereby certify that their views about the companies and their securities discussed in this report are accurately expressed and that they have not received and will not
receive direct or indirect compensation in exchange for expressing specific recommendations or views in this report: Stephen C Byrd; Carolyn L Campbell; James Egan; Catherine Liu; Vishwas
Patkar; Vishwanath Tirupattur; Lindsay A Tyler; Kyle B Yellin.
.

Global Research Conflict Management Policy

Morgan Stanley Research has been published in accordance with our conflict management policy, which is available at www.morganstanley.com/institutional/research/conflictpolicies. A
Portuguese version of the policy can be found at www.morganstanley.com.br

Important Regulatory Disclosures on Subject Companies

The equity research analysts or strategists principally responsible for the preparation of Morgan Stanley Research have received compensation based upon various factors, including quality
of research, investor client feedback, stock picking, competitive factors, firm revenues and overall investment banking revenues. Equity Research analysts' or strategists' compensation is not
linked to investment banking or capital markets transactions performed by Morgan Stanley or the profitability or revenues of particular trading desks.
Morgan Stanley and its affiliates do business that relates to companies/instruments covered in Morgan Stanley Research, including market making, providing liquidity, fund management,
commercial banking, extension of credit, investment services and investment banking. Morgan Stanley sells to and buys from customers the securities/instruments of companies covered in
Morgan Stanley Research on a principal basis. Morgan Stanley may have a position in the debt of the Company or instruments discussed in this report. Morgan Stanley trades or may trade
as principal in the debt securities (or in related derivatives) that are the subject of the debt research report.
Certain disclosures listed above are also for compliance with applicable regulations in non-US jurisdictions.

STOCK RATINGS
Morgan Stanley uses a relative rating system using terms such as Overweight, Equal-weight, Not-Rated or Underweight (see definitions below). Morgan Stanley does not assign ratings of Buy,
Hold or Sell to the stocks we cover. Overweight, Equal-weight, Not-Rated and Underweight are not the equivalent of buy, hold and sell. Investors should carefully read the definitions of all
ratings used in Morgan Stanley Research. In addition, since Morgan Stanley Research contains more complete information concerning the analyst's views, investors should carefully read Morgan
Stanley Research, in its entirety, and not infer the contents from the rating alone. In any case, ratings (or research) should not be used or relied upon as investment advice. An investor's decision
to buy or sell a stock should depend on individual circumstances (such as the investor's existing holdings) and other considerations.

Global Stock Ratings Distribution

(as of June 30, 2025)
The Stock Ratings described below apply to Morgan Stanley's Fundamental Equity Research and do not apply to Debt Research produced by the Firm.
For disclosure purposes only (in accordance with FINRA requirements), we include the category headings of Buy, Hold, and Sell alongside our ratings of Overweight, Equal-weight, Not-Rated
and Underweight. Morgan Stanley does not assign ratings of Buy, Hold or Sell to the stocks we cover. Overweight, Equal-weight, Not-Rated and Underweight are not the equivalent of buy,
hold, and sell but represent recommended relative weightings (see definitions below). To satisfy regulatory requirements, we correspond Overweight, our most positive stock rating, with a
buy recommendation; we correspond Equal-weight and Not-Rated to hold and Underweight to sell recommendations, respectively.

Morgan Stanley Research 35

M
North America Insight

Other Material Investment Services

Coverage Universe Investment Banking Clients (IBC)
Clients (MISC)

Stock Rating % of Rating % of Total Other

Count % of Total Count % of Total IBC Count
Category Category MISC

Overweight/Buy 1497 40% 372 45% 25% 702 41%

Equal-weight/Hold 1632 44% 380 46% 23% 776 45%

Not-Rated/Hold 4 0% 0 0% 0% 2 0%

Underweight/Sell 593 16% 79 10% 13% 233 14%

Total 3,726 831 1713

Data include common stock and ADRs currently assigned ratings. Investment Banking Clients are companies from whom Morgan Stanley received investment banking compensation in the
last 12 months. Due to rounding off of decimals, the percentages provided in the "% of total" column may not add up to exactly 100 percent.

Analyst Stock Ratings

Overweight (O). The stock's total return is expected to exceed the average total return of the analyst's industry (or industry team's) coverage universe, on a risk-adjusted basis, over the next
12-18 months.
Equal-weight (E). The stock's total return is expected to be in line with the average total return of the analyst's industry (or industry team's) coverage universe, on a risk-adjusted basis, over
the next 12-18 months.
Not-Rated (NR). Currently the analyst does not have adequate conviction about the stock's total return relative to the average total return of the analyst's industry (or industry team's) coverage
universe, on a risk-adjusted basis, over the next 12-18 months.
Underweight (U). The stock's total return is expected to be below the average total return of the analyst's industry (or industry team's) coverage universe, on a risk-adjusted basis, over the next
12-18 months.
Unless otherwise specified, the time frame for price targets included in Morgan Stanley Research is 12 to 18 months.

Analyst Industry Views

Attractive (A): The analyst expects the performance of his or her industry coverage universe over the next 12-18 months to be attractive vs. the relevant broad market benchmark, as indicated
below.
In-Line (I): The analyst expects the performance of his or her industry coverage universe over the next 12-18 months to be in line with the relevant broad market benchmark, as indicated below.
Cautious (C): The analyst views the performance of his or her industry coverage universe over the next 12-18 months with caution vs. the relevant broad market benchmark, as indicated below.
Benchmarks for each region are as follows: North America - S&P 500; Latin America - relevant MSCI country index or MSCI Latin America Index; Europe - MSCI Europe; Japan - TOPIX; Asia -
relevant MSCI country index or MSCI sub-regional index or MSCI AC Asia Pacific ex Japan Index.

Important Disclosures for Morgan Stanley Smith Barney LLC Customers

Important disclosures regarding the relationship between the companies that are the subject of Morgan Stanley Research and Morgan Stanley Smith Barney LLC or Morgan Stanley or any
of their affiliates, are available on the Morgan Stanley Wealth Management disclosure website at www.morganstanley.com/online/researchdisclosures. For Morgan Stanley specific disclosures,
you may refer to www.morganstanley.com/researchdisclosures.
Each Morgan Stanley research report is reviewed and approved on behalf of Morgan Stanley Smith Barney LLC. This review and approval is conducted by the same person who reviews the
research report on behalf of Morgan Stanley. This could create a conflict of interest.

Other Important Disclosures

Morgan Stanley Research policy is to update research reports as and when the Research Analyst and Research Management deem appropriate, based on developments with the issuer, the
sector, or the market that may have a material impact on the research views or opinions stated therein. In addition, certain Research publications are intended to be updated on a regular periodic
basis (weekly/monthly/quarterly/annual) and will ordinarily be updated with that frequency, unless the Research Analyst and Research Management determine that a different publication
schedule is appropriate based on current conditions.
Morgan Stanley is not acting as a municipal advisor and the opinions or views contained herein are not intended to be, and do not constitute, advice within the meaning of Section 975 of the
Dodd-Frank Wall Street Reform and Consumer Protection Act.
Morgan Stanley produces an equity research product called a "Tactical Idea." Views contained in a "Tactical Idea" on a particular stock may be contrary to the recommendations or views expressed
in research on the same stock. This may be the result of differing time horizons, methodologies, market events, or other factors. For all research available on a particular stock, please contact
your sales representative or go to Matrix at http://www.morganstanley.com/matrix.
Morgan Stanley Research is provided to our clients through our proprietary research portal on Matrix and also distributed electronically by Morgan Stanley to clients. Certain, but not all, Morgan
Stanley Research products are also made available to clients through third-party vendors or redistributed to clients through alternate electronic means as a convenience. For access to all
available Morgan Stanley Research, please contact your sales representative or go to Matrix at http://www.morganstanley.com/matrix.
Any access and/or use of Morgan Stanley Research is subject to Morgan Stanley's Terms of Use (http://www.morganstanley.com/terms.html). By accessing and/or using Morgan Stanley
Research, you are indicating that you have read and agree to be bound by our Terms of Use (http://www.morganstanley.com/terms.html). In addition you consent to Morgan Stanley processing
your personal data and using cookies in accordance with our Privacy Policy and our Global Cookies Policy (http://www.morganstanley.com/privacy_pledge.html), including for the purposes of
setting your preferences and to collect readership data so that we can deliver better and more personalized service and products to you. To find out more information about how Morgan Stanley
processes personal data, how we use cookies and how to reject cookies see our Privacy Policy and our Global Cookies Policy (http://www.morganstanley.com/privacy_pledge.html).
If you do not agree to our Terms of Use and/or if you do not wish to provide your consent to Morgan Stanley processing your personal data or using cookies please do not access our research.
Morgan Stanley Research does not provide individually tailored investment advice. Morgan Stanley Research has been prepared without regard to the circumstances and objectives of those
who receive it. Morgan Stanley recommends that investors independently evaluate particular investments and strategies, and encourages investors to seek the advice of a financial adviser.
The appropriateness of an investment or strategy will depend on an investor's circumstances and objectives. The securities, instruments, or strategies discussed in Morgan Stanley Research

36
M
North America Insight

may not be suitable for all investors, and certain investors may not be eligible to purchase or participate in some or all of them. Morgan Stanley Research is not an offer to buy or sell or the
solicitation of an offer to buy or sell any security/instrument or to participate in any particular trading strategy. The value of and income from your investments may vary because of changes
in interest rates, foreign exchange rates, default rates, prepayment rates, securities/instruments prices, market indexes, operational or financial conditions of companies or other factors. There
may be time limitations on the exercise of options or other rights in securities/instruments transactions. Past performance is not necessarily a guide to future performance. Estimates of future
performance are based on assumptions that may not be realized. If provided, and unless otherwise stated, the closing price on the cover page is that of the primary exchange for the subject
company's securities/instruments.
The fixed income research analysts, strategists or economists principally responsible for the preparation of Morgan Stanley Research have received compensation based upon various factors,
including quality, accuracy and value of research, firm profitability or revenues (which include fixed income trading and capital markets profitability or revenues), client feedback and competitive
factors. Fixed Income Research analysts', strategists' or economists' compensation is not linked to investment banking or capital markets transactions performed by Morgan Stanley or the
profitability or revenues of particular trading desks.
The "Important Regulatory Disclosures on Subject Companies" section in Morgan Stanley Research lists all companies mentioned where Morgan Stanley owns 1% or more of a class of common
equity securities of the companies. For all other companies mentioned in Morgan Stanley Research, Morgan Stanley may have an investment of less than 1% in securities/instruments or
derivatives of securities/instruments of companies and may trade them in ways different from those discussed in Morgan Stanley Research. Employees of Morgan Stanley not involved in the
preparation of Morgan Stanley Research may have investments in securities/instruments or derivatives of securities/instruments of companies mentioned and may trade them in ways different
from those discussed in Morgan Stanley Research. Derivatives may be issued by Morgan Stanley or associated persons.
With the exception of information regarding Morgan Stanley, Morgan Stanley Research is based on public information. Morgan Stanley makes every effort to use reliable, comprehensive
information, but we make no representation that it is accurate or complete. We have no obligation to tell you when opinions or information in Morgan Stanley Research change apart from
when we intend to discontinue equity research coverage of a subject company. Facts and views presented in Morgan Stanley Research have not been reviewed by, and may not reflect information
known to, professionals in other Morgan Stanley business areas, including investment banking personnel.
Morgan Stanley Research personnel may participate in company events such as site visits and are generally prohibited from accepting payment by the company of associated expenses unless
pre-approved by authorized members of Research management.
Morgan Stanley may make investment decisions that are inconsistent with the recommendations or views in this report.
To our readers based in Taiwan or trading in Taiwan securities/instruments: Information on securities/instruments that trade in Taiwan is distributed by Morgan Stanley Taiwan Limited ("MSTL").
Such information is for your reference only. The reader should independently evaluate the investment risks and is solely responsible for their investment decisions. Morgan Stanley Research
may not be distributed to the public media or quoted or used by the public media without the express written consent of Morgan Stanley. Any non-customer reader within the scope of Article
7-1 of the Taiwan Stock Exchange Recommendation Regulations accessing and/or receiving Morgan Stanley Research is not permitted to provide Morgan Stanley Research to any third party
(including but not limited to related parties, affiliated companies and any other third parties) or engage in any activities regarding Morgan Stanley Research which may create or give the
appearance of creating a conflict of interest. Information on securities/instruments that do not trade in Taiwan is for informational purposes only and is not to be construed as a recommendation
or a solicitation to trade in such securities/instruments. MSTL may not execute transactions for clients in these securities/instruments.
Morgan Stanley is not incorporated under PRC law and the research in relation to this report is conducted outside the PRC. Morgan Stanley Research does not constitute an offer to sell or
the solicitation of an offer to buy any securities in the PRC. PRC investors shall have the relevant qualifications to invest in such securities and shall be responsible for obtaining all relevant
approvals, licenses, verifications and/or registrations from the relevant governmental authorities themselves. Neither this report nor any part of it is intended as, or shall constitute, provision
of any consultancy or advisory service of securities investment as defined under PRC law. Such information is provided for your reference only.
Morgan Stanley Research is disseminated in Brazil by Morgan Stanley C.T.V.M. S.A. located at Av. Brigadeiro Faria Lima, 3600, 6th floor, São Paulo - SP, Brazil; and is regulated by the Comissão
de Valores Mobiliários; in Mexico by Morgan Stanley México, Casa de Bolsa, S.A. de C.V which is regulated by Comision Nacional Bancaria y de Valores. Paseo de los Tamarindos 90, Torre 1,
Col. Bosques de las Lomas Floor 29, 05120 Mexico City; in Japan by Morgan Stanley MUFG Securities Co., Ltd. and, for Commodities related research reports only, Morgan Stanley Capital Group
Japan Co., Ltd; in Hong Kong by Morgan Stanley Asia Limited (which accepts responsibility for its contents) and by Morgan Stanley Bank Asia Limited; in Singapore by Morgan Stanley Asia
(Singapore) Pte. (Registration number 199206298Z) and/or Morgan Stanley Asia (Singapore) Securities Pte Ltd (Registration number 200008434H), regulated by the Monetary Authority of
Singapore (which accepts legal responsibility for its contents and should be contacted with respect to any matters arising from, or in connection with, Morgan Stanley Research) and by Morgan
Stanley Bank Asia Limited, Singapore Branch (Registration number T14FC0118J); in Australia to "wholesale clients" within the meaning of the Australian Corporations Act by Morgan Stanley
Australia Limited A.B.N. 67 003 734 576, holder of Australian financial services license No. 233742, which accepts responsibility for its contents; in Australia to "wholesale clients" and "retail
clients" within the meaning of the Australian Corporations Act by Morgan Stanley Wealth Management Australia Pty Ltd (A.B.N. 19 009 145 555, holder of Australian financial services license
No. 240813, which accepts responsibility for its contents; in Korea by Morgan Stanley & Co International plc, Seoul Branch; in India by Morgan Stanley India Company Private Limited having
Corporate Identification No (CIN) U22990MH1998PTC115305, regulated by the Securities and Exchange Board of India (“SEBI”) and holder of licenses as a Research Analyst (SEBI Registration
No. INH000001105); Stock Broker (SEBI Stock Broker Registration No. INZ000244438), Merchant Banker (SEBI Registration No. INM000011203), and depository participant with National
Securities Depository Limited (SEBI Registration No. IN-DP-NSDL-567-2021) having registered office at Altimus, Level 39 & 40, Pandurang Budhkar Marg, Worli, Mumbai 400018, India;
Telephone no. +91-22-61181000; Compliance Officer Details: Mr. Tejarshi Hardas, Tel. No.: +91-22-61181000 or Email: tejarshi.hardas@morganstanley.com; Grievance officer details: Mr. Tejarshi
Hardas, Tel. No.: +91-22-61181000 or Email: msic-compliance@morganstanley.com. Morgan Stanley India Company Private Limited (MSICPL) may use AI tools in providing research services.
All recommendations contained herein are made by the duly qualified research analysts; in Canada by Morgan Stanley Canada Limited; in Germany and the European Economic Area where
required by Morgan Stanley Europe S.E., authorised and regulated by Bundesanstalt fuer Finanzdienstleistungsaufsicht (BaFin) under the reference number 149169; in the US by Morgan Stanley
& Co. LLC, which accepts responsibility for its contents. Morgan Stanley & Co. International plc, authorized by the Prudential Regulation Authority and regulated by the Financial Conduct
Authority and the Prudential Regulation Authority, disseminates in the UK research that it has prepared, and research which has been prepared by any of its affiliates, only to persons who (i)
are investment professionals falling within Article 19(5) of the Financial Services and Markets Act 2000 (Financial Promotion) Order 2005 (as amended, the “Order”); (ii) are persons who are
high net worth entities falling within Article 49(2)(a) to (d) of the Order; or (iii) are persons to whom an invitation or inducement to engage in investment activity (within the meaning of section
21 of the Financial Services and Markets Act 2000, as amended) may otherwise lawfully be communicated or caused to be communicated. RMB Morgan Stanley Proprietary Limited is a member
of the JSE Limited and A2X (Pty) Ltd. RMB Morgan Stanley Proprietary Limited is a joint venture owned equally by Morgan Stanley International Holdings Inc. and RMB Investment Advisory
(Proprietary) Limited, which is wholly owned by FirstRand Limited. The information in Morgan Stanley Research is being disseminated by Morgan Stanley Saudi Arabia, regulated by the Capital
Market Authority in the Kingdom of Saudi Arabia, and is directed at Sophisticated investors only.
The information in Morgan Stanley Research is being communicated by Morgan Stanley & Co. International plc (DIFC Branch), regulated by the Dubai Financial Services Authority (the DFSA)
or by Morgan Stanley & Co. International plc (ADGM Branch), regulated by the Financial Services Regulatory Authority Abu Dhabi (the FSRA), and is directed at Professional Clients only, as
defined by the DFSA or the FSRA, respectively. The financial products or financial services to which this research relates will only be made available to a customer who we are satisfied meets

Morgan Stanley Research 37

M
North America Insight

the regulatory criteria of a Professional Client. A distribution of the different MS Research ratings or recommendations, in percentage terms for Investments in each sector covered, is available
upon request from your sales representative.
The information in Morgan Stanley Research is being communicated by Morgan Stanley & Co. International plc (QFC Branch), regulated by the Qatar Financial Centre Regulatory Authority
(the QFCRA), and is directed at business customers and market counterparties only and is not intended for Retail Customers as defined by the QFCRA.
As required by the Capital Markets Board of Turkey, investment information, comments and recommendations stated here, are not within the scope of investment advisory activity. Investment
advisory service is provided exclusively to persons based on their risk and income preferences by the authorized firms. Comments and recommendations stated here are general in nature. These
opinions may not fit to your financial status, risk and return preferences. For this reason, to make an investment decision by relying solely to this information stated here may not bring about
outcomes that fit your expectations.
The trademarks and service marks contained in Morgan Stanley Research are the property of their respective owners. Third-party data providers make no warranties or representations relating
to the accuracy, completeness, or timeliness of the data they provide and shall not have liability for any damages relating to such data. The Global Industry Classification Standard (GICS) was
developed by and is the exclusive property of MSCI and S&P.
Morgan Stanley Research, or any portion thereof may not be reprinted, sold or redistributed without the written consent of Morgan Stanley.
Indicators and trackers referenced in Morgan Stanley Research may not be used as, or treated as, a benchmark under Regulation EU 2016/1011, or any other similar framework.
The issuers and/or fixed income products recommended or discussed in certain fixed income research reports may not be continuously followed. Accordingly, investors should regard those
fixed income research reports as providing stand-alone analysis and should not expect continuing analysis or additional reports relating to such issuers and/or individual fixed income products.
Morgan Stanley may hold, from time to time, material financial and commercial interests regarding the company subject to the Research report.

Registration granted by SEBI and certification from the National Institute of Securities Markets (NISM) in no way guarantee performance
of the intermediary or provide any assurance of returns to investors. Investment in securities market are subject to market risks. Read
all the related documents carefully before investing.
The following authors are Fixed Income Research Analysts/Strategists and are not opining on or expressing recommendations on equity securities: Carolyn L Campbell; Vishwanath Tirupattur;
James Egan; Vishwas Patkar; Lindsay A Tyler; Kyle B Yellin.

38
 © Morgan Stanley 2025

The Americas Europe Japan Asia/Pacific

1585 Broadway 20 Bank Street, Canary Wharf 1-9-7 Otemachi, Chiyoda-ku 1 Austin Road West
New York, NY 10036-8293 London E14 4AD Tokyo 100-8104 Kowloon
United States United Kingdom Japan Hong Kong
Tel: +1 (1) 212 761 4000 Tel: +44 (0) 20 7 425 8000 Tel: +81 (0) 3 6836 5000 Tel: +852 2848 5200