OpenAI to raise another $10 billion from funds for data centers and training new models

OpenAI plans to raise another $10 billion as part of its current funding round. Amazon, Nvidia, and SoftBank have already supplied most of the money, and the company needs the new tranche to scale up compute capacity faster and train the next generations of its language models.

Who is putting up the money

The new money would fill out the rest of a round that should bring OpenAI about $120 billion in total. According to industry sources, the company has already received the lion's share from Amazon, Nvidia, and SoftBank — players that benefit from staying close to the largest consumer of AI hardware and cloud resources. OpenAI is now closing out the tail end of the deal with other institutional participants: funds and large investment firms willing to buy into the company at a late stage.

These are not retail investors, and this is not a small bridge round. For the market, the tranche signals that demand for a stake in OpenAI remains high even after massive investments from strategic partners. The wider the pool of backers, the less OpenAI depends on any single group of funders, and the easier it becomes to build a long-term financial plan around expensive infrastructure rather than a single model release. It also strengthens OpenAI's hand in future negotiations over loans and new investment.

Where the money will go

OpenAI's biggest expense today isn't marketing or typical startup-style expansion — it's compute. The company burns billions training and running models, rents and reserves capacity, and signs long-term contracts with cloud providers and accelerator makers. It needs the new capital so it doesn't have to slow down upcoming model launches, while still handling the load from ChatGPT users, API customers, and enterprise clients, including for more powerful multimodal systems.

• expanding data centers and leasing new clusters
• buying and reserving accelerators for model training
• inference for ChatGPT, the API, and enterprise services
• developing new generations of language and multimodal models
• a liquidity cushion for long-term infrastructure contracts

These investments pay back over a long cycle: the money goes out now, and returns arrive later through subscriptions, enterprise licenses, and the API. As a result, OpenAI is effectively becoming not just a model developer but also the operator of a massive AI infrastructure. The faster its audience and query volume grow, the tighter the link becomes between model quality, hardware availability, and the cost of serving a single response. That's why the company tries to raise capital with room to spare rather than just enough to reach its next release.

Why the stakes keep rising

The $120 billion round shows how sharply the economics of generative AI have changed. Not long ago, big sums were raised mainly to hire teams and fund research. Now the main bottlenecks are energy, data centers, networking equipment, and chips. Investors aren't simply backing another chatbot; they're backing the foundational layer of future services. Whoever controls compute gains an edge both in how fast models get trained and in what it costs to run them for end customers.

For Amazon, Nvidia, and SoftBank, it's also a way to secure a foothold in the hottest segment of the AI market. For smaller institutional funds, joining a round like this is a chance to get into the company setting the pace for the entire industry. At the same time, it raises the bar for competitors: Anthropic, xAI, Google, and others have to think not only about model quality but also about where to find comparable compute and capital without losing flexibility.

What it means

If the deal closes in full, OpenAI will further cement its status as a company building not just a product but the infrastructure for the next stage of the AI market. For users, it means a chance to see new models and features sooner. For the industry, it's a reminder that the race increasingly goes to those with access to money, chips, and energy all at once. For smaller startups, it's also a signal that competing on an idea alone is no longer enough.