OpenAI's $110 billion round turns suppliers into shareholders

Amazon and Nvidia put in $80 billion and will sell compute back to OpenAI. Frontier valuations are now set by the vendors that power the labs.
OpenAI has raised $110 billion at a $730 billion pre-money valuation, nearly four times the $30 billion Anthropic had raised two weeks earlier. The round, announced on 27 February, rests on three checks: Amazon committed $50 billion, part of it payable once preset conditions are met, Nvidia $30 billion and a financial investor another $30 billion.
Close to three quarters of that money comes from companies that will sell OpenAI the chips and cloud capacity it runs on, and that changes who sets the price of a frontier lab. Strategic buyers of demand now set it, rather than financial investors pricing cash flows. Amazon and Nvidia are paying for a customer's growth and can expect much of the money to return to them as revenue, which makes the valuation partly a commercial decision, as Amazon's Anthropic stake did in 2023.
The compute attached to both strategic checks makes the loop explicit. OpenAI's existing cloud agreement with Amazon grows by $100 billion over eight years, and its Nvidia partnership covers 3 gigawatts of dedicated inference capacity and 2 gigawatts of training on Vera Rubin systems. Arrangements like these support very high valuations at the top for as long as they last, and they tie the health of the largest cloud and chip businesses to a small number of labs. Every investor in the AI stack now carries that concentration, the theme of our 2025 review.
The frontier has become a contest of balance sheets that few companies can enter, so new companies win elsewhere. The open ground lies in applications and infrastructure that make the labs' models useful inside businesses, where capital needs are modest and margins depend on product, not on gigawatts. Distribution is shifting too. AWS becomes the exclusive third-party cloud for OpenAI Frontier, the company's new platform for running agents inside large enterprises, so a startup selling to the same buyers should expect clouds to bundle model access with their own sales channels. Partnerships with those channels are worth planning early rather than after the first enterprise deal.
The headline number is softer than it looks. The terms behind Amazon's conditional tranche are not public, and whether that money arrives on schedule remains open. Whether other investors join before the round closes, and at what price, will test demand from investors who do not sell compute. Over the coming quarters the cloud and data center results of the new shareholders will show how much of OpenAI's spending flows back to them.
Frontier labs are becoming infrastructure companies financed by their own vendors. That calls for discipline on price at the model layer, and it is why Fanaura keeps its patience for the layers above, where a smaller check still buys a real share of the value created.
Building an AI-native company? Apply in five minutes. →
More from Fanaura
Building the application layer that turns frontier models into enterprise results? Send us the project.
Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.
Get the news by email
One short note each weekday. No marketing, and you can unsubscribe at any time.