Sam Altman on OpenAI's $850 billion in planned buildouts: 'People are worried. I totally get that' (www.cnbc.com)

OpenAI announced a massive infrastructure push: a 17‑gigawatt, multi‑site data center buildout in partnership with Oracle, Nvidia and SoftBank that OpenAI says totals roughly $850 billion in commitments (about $50 billion per large site), starting with a leased facility in Abilene, Texas. CEO Sam Altman framed the program as a necessary response to explosive demand—ChatGPT usage reportedly rose tenfold in 18 months—and said the biggest constraint is electricity, not chips. To secure steady baseload power, Altman has invested in nuclear and fusion firms (Helion, Oklo), while Nvidia is providing accelerators (including its Vera Rubin class) and equity up front; Oracle will lease and operate facilities and OpenAI will cover operating expenses once online. For the AI community this is consequential on multiple fronts: it aims to relieve an acute compute crunch and enable larger, more frequent training and inference workloads, but it also concentrates power (and financial flows) around a few vendors and creates large energy and capital risks. Critics call it potentially circular financing and worry about an investment bubble; proponents argue coordinated, large‑scale buildouts are historically required for new tech waves. Timelines point to capacity arriving next year with further expansion through 2027–29. The program also signals OpenAI’s long‑term push into hardware and devices (notably the Jony Ive acquisition), which could change how AI services map onto consumer endpoints.