OpenAI, Oracle, SoftBank, and investment firm MGX on Tuesday announced plans to spend $500 billion on AI infrastructure in America over the next four years, to be used by the ChatGPT super-lab.
Specifically, the four will pour said billions, starting with $100 billion, into a new company named The Stargate Project, which will be operated by OpenAI. Stargate will effectively take that money and build AI infrastructure for OpenAI to use to create more machine-learning models.
Financial management will be handled by Japan's SoftBank, which in December committed $100 billion in US investment over the next four years; SoftBank supremo Masayoshi Son will be chairman of Stargate.
"We will begin deploying $100 billion immediately," the companies wrote in a joint statement. "This infrastructure will secure American leadership in AI, create hundreds of thousands of American jobs, and generate massive economic benefit for the entire world."
At a White House briefing where the plan was announced, Oracle chairman and Republican Larry Ellison said: "The data centers are actually under construction. The first of them are under construction in Texas. Each building is half a million square feet. There are 10 buildings currently being built but that will expand to 20, and other locations beyond the Abilene [Texas] location, which is our first location."
Texas is also among the states cited by Emirati luxury property developer DAMAC, which recently pledged to invest $20 billion to support the construction of datacenters in the US.
Other build sites elsewhere in America are also under consideration. Given OpenAI's reported talks with Broadcom, some of these sites may eventually house silicon designed specifically for OpenAI.
The deal conveniently follows the Trump administration rescinding Executive Order 14110, the Biden administration's order on Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
The Stargate Project involves SoftBank-owned Arm, Microsoft, Nvidia, Oracle, and OpenAI as technology partners. It represents a continuation of OpenAI's collaboration with Nvidia and adds Oracle as a new collaborator.
Oracle and Nvidia shares rose in after-hours trading following Stargate's announcement. Microsoft shares, perhaps in response to a commitment by OpenAI "to increase its consumption of Azure," also saw after-hours gains.
That commitment, by the way, came in a carefully worded announcement in which Microsoft - which has poured billions into OpenAI to use the super-lab's models to power its own Copilot family of AI assistants - revealed it was modifying its 10-year deal with OpenAI so that the Windows giant isn't necessarily the lab's sole exclusive cloud provider going forward.
Microsoft will have a right to first refusal for providing OpenAI with more compute capacity as the latter needs, and if that doesn't work out, OpenAI can turn to another provider. It sound as though OpenAI intends to score more compute capacity for carrying out research and model training.
Back to Stargate, and as an example of the kinds of applications the project expects to run on its shiny new infrastructure, Ellison cited electronic health record management that would help doctors employ AI to treat patients.
Softbank's Masayoshi Son endorsed that sort of outcome, saying "...artificial super intelligence will come to solve the issues that mankind would never ever have thought that we could solve. This is the beginning of our golden age."
"All of us look forward to continuing to build and develop AI - and in particular AGI - for the benefit of all of humanity," the companies said in their statement, referring to "artificial general intelligence," an ill-defined notion of AI models capable of human-level cognition that OpenAI CEO Sam Altman only a day earlier sought to downplay. "We believe that this new step is critical on the path, and will enable creative people to figure out how to use AI to elevate humanity."
If Stargate manages to spend $100 billion on AI infrastructure in 2025 alone, it will top Microsoft's planned $80 billion building spree and probably beat Amazon's planned expenditure which has been forecast to top the $75 billion the outfit spent in 2024. Google has told investors that for 2025 it plans to blow through more than the $50 billion-plus it likely spent on compute infrastructure in 2024, though that increase (in percentage terms) will be less than the 60 percent jump from 2023 to 2024.
Microsoft recently indicated how its spending spree on infrastructure will be paid for by hiking the price of some of its 365 bundles up to 45 percent.
Maybe you can pay for that by selling some shares? ®
Justin Hotard tapped to replace Pekka Lundmark at the Finnish telco
'Intense year' ahead, warned Zuck. Got to spend billions on AI and work to stay out of Trump's bad books
Opinion When your state machines are vulnerable, all bets are off
FOSDEM 2025 OKD project also has its own immutable CentOS image, which could be fun
Costs for fixing them and keeping them working up by 390%, NAO report reveals
Also claims it's found DeepSeek-eque optimizations that reduce AI infrastructure requirements
Open source project chief hits out at 'social media brigading'