Plow $500 billion! NVIDIA teams up with OpenAI, with plans to build the “largest” AI data center worldwide, as Japan pours a huge amount of money into the deal

According to Forbes, NVIDIA and OpenAI are in intensive discussions about a super data center project worth as much as $500 billion. The facility is expected to be located in the U.S. state of Ohio, with a designed power capacity of up to 10 gigawatts (GW), and could become the largest-scale AI infrastructure in the world. The deal also has geopolitical implications for the U.S.-Japan region: Japan has pledged to invest $33 billion in exchange for tariff relief.
(Background: Huang Renxun boldly supports China’s AI! He says the U.S. is stuck in “Kimi panic” and “getting the wrong focus”—free open source will unleash NVIDIA chip demand)
(Additional context: NVIDIA launches 14 kinds of “GeForce physical transaction cards”! Featuring legendary GPUs from the past—Summer of RTX free drawing)

Table of Contents

Toggle

• Build a 10 GW computing powerhouse with $500B
• Japan enters with heavy investment, with the U.S. commerce minister personally steering the decision
• Tech giants spend in the trillions to expand AI infrastructure

Amid the increasingly hot AI arms race, the scale of compute infrastructure is being pushed to unprecedented heights. On July 27, Taipei time, according to a report by Forbes reporter Ty Roush citing The Wall Street Journal, NVIDIA is exploring a partnership deal with ChatGPT developer OpenAI that would shock the industry. The two sides plan to jointly build a super data center in southern Ohio—about 50 miles south of Columbus—with a power capacity of up to 10 gigawatts (GW).

Build a 10 GW computing powerhouse with $500B

According to the report, NVIDIA is expected to provide about $250 billion in funding support for the project, helping OpenAI rent the proposed site developed by SoftBank’s energy subsidiary SBEnergy. If the hardware costs of NVIDIA AI chips are included, the total investment for the entire project could exceed $500 billion; meanwhile, OpenAI is also concurrently negotiating a chip procurement deal worth as much as $350 billion.

How big is this 10 GW data center? Data shows it could supply electricity for as many as 8.4 million typical U.S. households. Compared with the $20 billion, 3.2 GW project that OpenAI just announced in Georgia last week, the scale of the Ohio plan is undoubtedly an unprecedented super giant. The official timeline indicates that the facility’s first-phase construction is expected to be completed in 2028, at which point it will initially provide about 800 megawatts (MW) of power.

Japan enters with heavy investment, with the U.S. commerce minister personally steering the decision

Notably, the data center would be located on U.S. federal land and has strong geopolitical overtones. In exchange for lower tariffs on U.S. exports, Japan has promised to inject $33 billion into the project. Under the agreement between the two sides, the U.S. and Japan would share the initial power-sales revenue until Japan fully recoups its $33 billion investment principal, after which the U.S. would hold absolute control of 90% of the facility.

Due to the project’s high national security and strategic value, U.S. Commerce Secretary Howard Lutnick is also reportedly personally involved in the decision-making—responsible for deciding which companies ultimately get access to the valuable power resources at the site. In addition to OpenAI, including Microsoft, Google, and the startup heavyweight Anthropic, all have expressed strong interest in moving in to Lutnick.

Tech giants spend in the trillions to expand AI infrastructure

This super project, still under discussion, highlights this year’s trend of massive capital spending on AI among major technology companies. Looking back at recent market dynamics, tech giants’ investments in infrastructure have become almost frantic: Meta plans to expand its data centers in Louisiana to 5 GW (total investment of more than $50 billion); Alphabet has just secured a $10 billion injection from institutions including Berkshire Hathaway to build AI facilities; Amazon has also increased its investment in Anthropic to $25 billion, while Anthropic has promised that over the next 10 years it will purchase more than $100 billion in services from Amazon Web Services.

As an “AI arms dealer,” NVIDIA only announced on Friday that it has reached an AI memory supply transaction with SK Hynix with a total value of more than $500 billion (including large data center construction expected to go live in 2027). With major players continuing to throw money into the effort, the global competition for AI compute infrastructure is clearly only just entering its most critical final battle phase.