The Blog
Engineering notes, scaling stories, and the occasional rant — straight from the team building the cloud for the AI era.
Release 11 “Metis”: Snapshots, Sandboxes, and a Shield
Branching and checkpointing, persistent custom filesystems, a plugin API, and a network shield that keeps secrets out of your microVMs entirely — plus a completely revamped dashboard. Here is everything in R11.
Felipe Huici
Co-Founder & CEO
The Great Fragmentation of Cloud Compute
Why does a single cloud sell you four compute services for one workload — and what would it take to defragment them back into one?
The Big AI Divide
AI collapsed the cost of building a virtualization platform, but running one in production at scale is just as hard as ever. Here's where the real work now lives.
Say No to Warm Pools
You provision always-on clusters, provisioned for a spike that spends most of its life not happening. Here is what they actually cost you — and why millisecond cold starts make the whole pattern obsolete.
Why do I need Unikraft? I'll just grab Firecracker
Firecracker is a superb virtual machine monitor, but it is only one component of a full platform. Here is exactly what you get when you run Unikraft instead of building it yourself.
Release 11 “Metis”: Snapshots, Sandboxes, and a Shield
Branching and checkpointing, persistent custom filesystems, a plugin API, and a network shield that keeps secrets out of your microVMs entirely — plus a completely revamped dashboard. Here is everything in R11.
The Mighty (micro)Virtual Machine: The Only Isolation Primitive That Actually Isolates
A no-BS guide to VMs, containers, microVMs, and isolates — and why, for multi-tenant workloads, there is, and only ever was, one right answer.
Cramming 1M (Scaled to Zero) Virtual Machines in a Single Box
At Unikraft, we're on a quest for exponentially better cloud efficiency. We redesigned our platform to fit over 1 million scaled-to-zero microVMs on a single standard server, with cold starts and resumes still in single-digit milliseconds.
AI Startup? You Need to Think about Scale from Day 1
The era of "move fast, fix infra later" is over for AI companies. It used to be the case that if you were an early stage startup the name of the game would be to get whatever infra would get you started the quickest. That appears to no longer be the case.
We Raised $6M to Build the Future of (AI) Cloud Infrastructure
Today, and after years of hard work, we’re ecstatic to announce the launch of Unikraft Cloud, a radical, new cloud platform based on years of research and providing exponentially better scalability and efficiency (think running millions of strongly isolated instances on a few servers instead of an entire datacenter). In support of this, we’re also super proud to announce our latest $6M seed raise with Heavybit as lead investor, with the participation of Vercel Ventures, Mango Capital, Firestreak Ventures, Fly VC and First Momentum Ventures, along with a host of world class angels.
Running Thousands of Millisecond Headless Browsers on a Single (Cloud) Server
Headless browsers have become a fundamental cog in the AI machine, running as a cloud service for running JavaScript code, scraping websites, automating testing, and integrating with AI agents. A headless browser is, however, a resource-hungry, heavy beast. In this article we show how Unikraft Cloud tech allows for deploying 1000s of these on a single server, while ensuring they can be started in 10s of milliseconds.
SOC2-I Compliance: An Intense but Extremely Valuable Journey
At Unikraft, we recently completed our SOC2 Type 1 audit. What we thought would be a process of improving security turned into a much more valuable journey about company maturity.
The Idle (and Over-provisioned) Cloud
We generally like to think of the cloud handling gazillion requests per second, but a significant part of it is over-provisioned and so idle. In this post we look into the what, how and why what we term "The Idle Cloud" exists, and how to fix it.