'Stargate' is now a real thing, but sadly not a portal to an alien planet: A bunch of tech companies plan to spend $500 billion building AI data centers

OpenAI and SoftBank are leading a joint venture that plans to spend $500 billion on AI infrastructure in the United States. The project was underway before today, but the announcement was saved for President Trump's first week back in office—the president was joined on Tuesday by OpenAI founder Sam Altman, Softbank CEO Masayoshi Son, and Oracle founder Larry Ellison to talk up the new company.

The venture is called Stargate, but has nothing to do with an Egyptian wormhole, I'm afraid. It's being run by OpenAI and Softbank, and the goal is to spend $500 billion to build AI infrastructure in the US over four years, with $100 billion being deployed now. The first facility is already being built in Texas, OpenAI said in a statement, and the ChatGPT company is "evaluating potential sites across the country for more campuses." Trump claims the project will create 100,000 American jobs "almost immediately."

The president has also revoked a Biden order which aimed to put guardrails on AI development. The order wasn't especially effective according to a few accounts I've seen, but its dismissal sends the message that AI developers have at least four years to do what they want in the US without the threat of federal intervention.

Along with OpenAI and Softbank, Oracle and AI investor MGX have put cash into the Stargate pot, and the project's technology partners include Nvidia, Microsoft (which has invested over $10 billion in OpenAI), and Arm. As for what this infrastructure is, exactly, Trump said that they're building "colossal data centers."

That the AI industry is investing in data centers is nothing new—Google's even looking at mini nuclear reactors as power sources—though we don't know exactly what OpenAI and friends plan to do with these new facilities. The most specific example came from Ellison, who said that they'll use the infrastructure to analyze health records and help doctors make diagnoses. Later in the press conference, Altman said that he believes diseases, including cancers, will be cured at "an unprecedented rate" thanks to AI development, and Ellison added that they're working on early cancer detection and claimed they'll be able to develop individually-tailored cancer vaccines. (I found a company called Evaxion that says it is working on technology like this.)

Machine learning includes many applications—self-driving cars, language translation, Nvidia's latest rendering tech, and so on—but OpenAI itself has been particularly interested in generating text, images, and video with consumer tools like ChatGPT, and claims to be on the road to developing "artificial general intelligence," a hypothesized system that can solve problems with human-like versatility. Son claimed at the press conference that AGI is coming soon, and superintelligence after that.

Skeptics question whether a large language model can ever achieve AGI, and the biggest AI skeptics question whether chatbots and image/video generators are useful at all, as they're prone to error and misuse and generally rely on ingesting large amounts of copyrighted material without permission. The biggest believers in the potential of large language models, meanwhile, say that we're on the verge of a technological revolution so momentous that it could make money obsolete. OpenAI itself has cautioned investors that "it may be difficult to know what role money will play in a post-AGI world."

The biggest gaming news, reviews and hardware deals

Keep up to date with the most important stories and the best deals, as picked by the PC Gamer team.

Contact me with news and offers from other Future brandsReceive email from us on behalf of our trusted partners or sponsors

OpenAI was originally a non-profit company, but now controls a second, capped for-profit company. Ellison and Son can both be found on Bloomberg's list of the 500 richest people in the world , but with an estimated net worth of $2 billion , Altman doesn't make the cut.