OpenAI and Anthropic are driving hundreds of billions of dollars in AI infrastructure investment, but they are not necessarily the companies paying upfront to build the data centers running t
OpenAI and Anthropic are driving hundreds of billions of dollars in AI infrastructure investment, but they are not necessarily the companies paying upfront to build the data centers running their models.
Instead, much of the construction is financed by cloud providers, infrastructure developers, banks, private-credit funds and institutional investors.
OpenAI and Anthropic often become long-term customers, committing to pay for computing capacity over many years. Those contracts help developers borrow billions before facilities begin generating revenue.
The result is a financing system where the company using a data center, the company owning it and the investors funding it can be entirely different entities.
Who Actually Funds OpenAI's Data Centers?Oracle is one of OpenAI's major infrastructure partners.
In February 2026, Oracle announced plans to raise $45 billion–$50 billion through debt and equity to expand cloud infrastructure for customers including OpenAI.
That means Oracle and its investors provide capital to build computing capacity, while OpenAI's future payments help support the investment.
A separate Stargate project in Michigan illustrates another structure. The development involves Related Digital, Oracle, OpenAI and Blackstone, with financing from infrastructure investors and long-term debt backed by PIMCO-managed funds.
OpenAI does not necessarily own the buildings or carry the construction debt directly.
Anthropic's Infrastructure Is Financed DifferentlyAnthropic relies heavily on Amazon, Google and specialized data center operators.
Its Amazon agreement includes commitments exceeding $100 billion over ten years, securing access to as much as five gigawatts of computing capacity.
Meanwhile, TeraWulf signed a 20-year lease with Anthropic for a Kentucky data center, representing approximately $19 billion in contracted revenue.
The developer owns and builds the infrastructure, while Anthropic commits to payments over the lease term.
Chip financing adds another layer.
Broadcom has been linked to a potential $42 billion financing arrangement supporting Anthropic's computing requirements, showing how hardware suppliers could also become lenders.
That structure is part of a broader trend toward hidden AI liabilities.
Note: Figures represent different types of financing and contractual commitments, not directly comparable construction costs. Broadcom financing remains potential.
Who Loses Money If AI Demand Disappoints?The answer depends on the contracts.
If OpenAI or Anthropic fails to generate sufficient revenue, the financial consequences could spread across several groups.
Data center owners may struggle to collect rent. Banks and bondholders could face losses on loans. Equity investors could see project valuations decline.
However, long-term leases and guarantees may require customers or other counterparties to keep paying even when computing demand weakens.
That is why investors increasingly examine Anthropic's compute costs and the financing structures behind AI expansion.
The central distinction is simple: OpenAI and Anthropic create much of the demand, but Wall Street and infrastructure companies often provide the upfront money.
Ultimately, AI customers must generate enough revenue to support those payments, or the financial risk could spread far beyond the companies developing the models.
