Sign up to see the future, today<\/p>\n
Can\u2019t-miss innovations from the bleeding edge of science and tech<\/p>\n
Despite promising to allocate hundreds of billions of dollars for the build out of enormous data centers, the AI industry has struggled to keep up with its lofty ambitions. <\/p>\n
According to recent reporting by Bloomberg<\/a> and Ed Zitron<\/a>, roughly half the data centers slated to open in the United States are either being delayed or canceled outright. Massive electric component shortages and soaring costs have slowed the infrastructure boom to a trickle, frustrating tech leaders.<\/p>\n In the resulting morass, every AI player has been making big claims to try to keep the hype train going. But OpenAI in particular has turned braggadocio into an art form, and its latest boast is ringing a bit sad: in a memo obtained by Bloomberg<\/a>, the company boasted that it\u2019s planning to have 30 gigawatts worth of compute \u2014 enough to power over 22 million US households \u2014 by 2030, while its rival Anthropic is only planning for seven to eight gigawatts by the end of 2027.<\/p>\n To put those numbers into perspective, OpenAI had just 1.9 gigawatts of computing capacity in 2025, while Anthropic had 1.4 gigawatts.<\/p>\n \u201cEven at the high end of that range, our ramp is materially ahead and widening,\u201d OpenAI\u2019s memo reads.<\/p>\n The ChatGPT maker also argued that it was outpacing Anthropic by \u201crapidly and consistently\u201d adding computing capacity.<\/p>\n \u201cThat gap matters because compute is now a product constraint,\u201d the memo reads.<\/p>\n In other words, OpenAI\u2019s big brag isn\u2019t some amazing new technical breakthrough, perhaps one that could achieve impressive AI with far less computing power. Instead, it\u2019s just that it\u2019s building mountains of data centers to overwhelm the competition through brute force.<\/p>\n The timing of OpenAI\u2019s memo is telling. It was sent out just after Anthropic showed off its latest AI model, Claude Mythos, which its staffers allege<\/a> is too powerful and therefore too much of a risk to cybersecurity<\/a> to be released in full.<\/p>\n In a statement to Bloomberg, Anthropic struck back at OpenAI, pointing to a recent deal the company had struck with Broadcom and Google as part of \u201cour disciplined approach to scaling infrastructure.\u201d<\/p>\n \u201cWe are making our most significant compute commitment to date to keep pace with this unprecedented growth,\u201d the statement reads.<\/p>\n OpenAI, on the other hand, has been far more brash in its plans to dominate the industry, claiming it will spend a whopping $600 billion on AI infrastructure through 2030 \u2014 which, it\u2019s worth noting, is less than half<\/a> of what it originally promised to spend.<\/p>\n It\u2019s a precarious point of inflection for the Sam Altman-led company. As the walls continue to close in and investors become increasingly antsy ahead of its rumored blockbuster IPO, OpenAI has consolidated its plans to pursue many of the same goals<\/a> as Anthropic.<\/p>\n The central premise: the more compute, the more powerful the AI.<\/p>\n \u201cEach new generation of infrastructure lets us train more capable models, making every token more intelligent than the one before,\u201d OpenAI wrote in its memo. \u201cAt the same time, algorithmic gains and hardware improvements reduce the cost to serve each token, lowering the cost per unit of intelligence.\u201d<\/p>\n