Elon Musk unveils xAI plan for 50M H100-scale units as OpenAI and Oracle expand Stargate beyond 5GW to lead the AI infrastructure race.
Oracle and OpenAI have announced a 4.5 gigawatt (GW) expansion to facilitate future AI development.
OpenAI announced on Tuesday that the agreement, which is a component of its long-term strategy to deploy 10 gigawatts of compute capacity throughout the United States, will augment its current Stargate I facility in Abilene, Texas, and exceed the project’s initial commitment at the White House in January.
OpenAI CEO Sam Altman shared images of the Abilene site on X, stating, “This is a gigantic infrastructure project.” In an earlier post, he verified that over one million GPUs would be online by the end of the year and jokingly stated, “Now they better get to work figuring out how to 100x that.”
With the inclusion of the Oracle agreement, Stargate’s total development pipeline will exceed 5 GW, sufficient to power more than two million AI chips. In addition, Altman stated, “We intend to substantially increase the scope of Stargate’s objectives beyond the $500 billion commitment we made in January.”
Musk discloses an ambitious strategy for xAI.
Elon Musk also disclosed an ambitious strategy for his AI company, xAI, after OpenAI’s announcement. In a Tuesday post on X, the entrepreneur stated that the @xAI objective is to bring 50 million units of H100 equivalent-AI compute (with significantly improved power efficiency) online within five years.
TeslaPrice, an X user, has estimated that this would be equivalent to 500 times the compute capacity of the world’s most powerful AI supercomputer just one year ago.
550,000 GB200 processors, which are approximately equivalent to 5.5 million H100s, will be employed by XAI’s Colossus 2 supercomputer, which is scheduled to be operational shortly. If executed, Musk’s intentions regarding xAI would exceed that figure by nearly a factor of 10.
“Elon claims that they will reach the equivalent of 50 million H100 within five years.” So that will be 500 times more than the state-of-the-art cluster from 12 months ago,” the user estimated.
The $500 billion Stargate AI initiative is confronted with obstacles.
A $500 billion AI infrastructure initiative headed by the private sector, Stargate, was announced by US President Donald Trump earlier this year.
The initiative, which Oracle, OpenAI, and SoftBank support, is designed to establish AI data centers throughout the United States and generate more than 100,000 employment opportunities.
Nevertheless, the initiative has encountered significant delays and internal disagreements between key partners SoftBank and OpenAI, per a recent report from The Wall Street Journal.
According to the report, the project has reduced its near-term objectives to constructing a single data center by the end of the year, despite its initial commitment to deploy $100 billion immediately.
