kind of incredible that if you bought stock in #MicroStrategy¹ / #MSTR 25 years ago, well before it turned itself into the 500 lb gorilla of the “borrow money to buy bitcoin” industry but only shortly before Michael #Saylor's first² #dotcom era accounting fraud was uncovered and he set a world record for “most money lost in a single day”, you’d currently still have lost money.

¹ sorry i just can’t call it the new name (“Strategy”) with a straight face

² there’s been a few #michaelsaylor accounting frauds at this point, most recently a $50 million tax fraud settlement last year. almost makes you think this guy might not be on the up and up.

Alt...

chart showing that the price of MSTR is now below the maximum price it reached in the dotcom bubble

All it'd take is one clever math result demonstrating you don't need absolutely gigantic neural networks trained on mind-bogglingly-huge datasets to achieve the AI goals of most companies, and NVIDIA's hardware dominance evaporates. Why would you spend thousands or tens of thousands of dollars on a GPU that uses 300 Watts of power when you could achieve the same thing with an ASIC or FPGA that uses 3 Watts? This is already true for many applications, but apparently it hasn't been widely realized yet. It'll be hard to ignore if/when it becomes true for the vast majority of applications. Which it could.

#AI #LLM #hype #NVIDIA #graphics #bubble #DotCom