Oracle, OpenAI, and investors in Japan and the UAE have launched a $100 billion effort to build data centers to run AI applications, an indication of how the U.S.-China race for artificial intelligence is beginning to turn on sheer computing power instead of clever programming.
Oracle has already begun building 10 data centers in Texas of a half-million square feet apiece for the project, dubbed Stargate, CEO Larry Ellison said Tuesday at a White House press conference.
“That will expand to 20, and other locations,” Ellison said.
The venture is being operated by generative AI company OpenAI, which will be the primary user of the data centers, a person familiar with the deal told Defense One.
“I think this will be the most important project of this era,” OpenAI founder Sam Altman said at the press conference. “For [artificial general intelligence] to get built here, to create hundreds of thousands of jobs, to create a new industry centered here.”
The primary investors are private companies, including Japanese holding company SoftBank and United Arab Emirates-based MGX. Their initial investment will be $100 billion; Stargate aims to draw $500 billion within five years.
The press conference also featured President Donald Trump, who took credit for organizing the project but did not say what role the U.S. government might be playing in Stargate.
“What we want to do is, we want to keep it in this country. China is a competitor,” Trump said.
More about stuff, less about math
U.S. officials across the U.S. political spectrum have emphasized the importance of retaining the U.S. lead in AI over China. Last week, the outgoing Biden-administration national security advisor, Jake Sullivan, predicted “catastrophe” if the U.S. can’t retain its market and research advantage and lead the development of global norms around AI use.
In an interview with Axios, Sullivan said the consequences would be “dramatic, and dramatically negative—to include the democratization of extremely powerful and lethal weapons; massive disruption and dislocation of jobs; an avalanche of misinformation.”
In 2017, China released a plan to become a “world leader” in AI by 2050. Five years later, the Special Competitive Studies Project said the U.S. retained a narrow lead in AI research. China has since tried to accelerate its efforts; last August, the Chinese government announced a 43.5 billion yuan (US$6.1 billion) effort to build eight new computing hubs.
On Tuesday, SCSP President Ylli Bajraktari told Defense One, “While the U.S. has clear advantages, China is rapidly catching up. Namely, China has proven adept at deploying large-scale infrastructure projects quickly. So the details of this announcement will be crucial, but it’s a significant strategic step in the right direction.”
Bajraktari said the AI race is changing, becoming less about new mathematical models and much more about the hard physics of power and data storage.
“The biggest bottleneck to AI advancement in the United States isn’t a lack of innovation, but a lack of infrastructure. This includes not just data centers, but the energy grids and high-speed networks needed to support them,” he said.
That view was echoed by Greg Allen, who leads the Wadhwani AI Center at the Center for Strategic and International Studies.
“The emphasis of strategic competition in AI is moving away from software engineering secrets that countries try to protect and toward massive scale hardware AI infrastructure that countries try to build and deploy,” Allen said.The White House announcement is a good sign, said Daniel Castro, vice president at the Information Technology and Innovation Foundation.
“This partnership reflects a shared goal: creating the foundation for the U.S. to thrive in the AI economy. While the global AI race won’t be won overnight, scoring major points on day one of this new administration shows the U.S. tech industry is reinvigorated and ready to take bold steps forward with this new administration,” Castro said.
Yet the Biden administration also worked hard to foster the U.S. lead in AI. It created a $30 million National AI Research Resource pilot last March, proposed a $3 billion boost to federal use of AI, and installed export controls to slow China’s access to new AI-relevant computer chips. The Biden White House also issued an executive order on AI risks and a national security memorandum on AI.
On Monday, Trump rescinded the executive order, but CSIS’ Allen suggested not making too much of the move.
“Much of the directed federal government work of the Biden administration’s AI executive order has already been completed on schedule,” he said, adding: I think it is noteworthy that President Trump did not rescind the AI national-security memorandum.”
