All news

OpenAI's cloud deal makes compute the currency of AI funding

A brass key with a microchip for a bow, lying on dark blue velvet

OpenAI took $1 billion from a cloud provider that will also supply its computing. For AI labs, capital and compute now arrive as one package.

OpenAI has agreed to take $1 billion from one of the largest cloud providers, four months after it created a capped-profit company to raise outside capital. Under the multiyear deal, signed on 22 July, OpenAI will move its services onto the investor's cloud, which becomes its exclusive computing partner, and the two will jointly build new AI supercomputing systems.

At the research frontier, capital and compute now arrive as one package, and that changes who can lead a round. In March OpenAI said it would need to invest billions of dollars in the coming years in cloud compute, talent and AI supercomputers. This deal shows how that bill gets paid. The investor is also the supplier, so the money OpenAI spends on compute flows back to the investor's own data centers, and the lab gets capacity on terms no ordinary venture round could offer.

The rest of the structure points the same way. OpenAI will license some of the technologies it develops along the way, with the investor as its preferred partner for bringing them to market. Returns for the first investors are capped at 100 times their stake under the March structure, and profit above the cap flows to the nonprofit that controls the company.

For Fanaura, a family office investing in AI-first companies, the lesson is about pricing: a round of this kind is measured in processor hours as much as in dollars, and its valuation reflects a commercial contract as well as a stake. Financial investors with no cloud to offer will find it harder to lead rounds in labs like this one.

Few companies will raise from a cloud on these terms, but the template reaches well beyond one lab. The clouds now compete to host machine learning work, and credits, discounted capacity and joint selling are on the table for start-ups with demanding workloads. Those offers deserve the same scrutiny as the equity, since exclusive infrastructure, commercialization rights and preferred-partner clauses shape every later round and exit. A company that keeps its models portable across clouds keeps its negotiating power, a point we raise early when we work with founders on their first rounds.

Skeptics will say that one lab with an unusual charter proves little about the market. The evidence will come quickly. If the other large clouds sign similar pairings with independent research labs, the model spreads. The first technology OpenAI licenses under the deal will show whether a capped-profit lab can earn revenue of its own, and the speed at which it spends the capital will show what it builds first, a larger language model or a successor to the game-playing system that beat the Dota 2 world champions.

Frontier machine learning research is becoming a capital-intensive business tied to the largest clouds. Financial investors will find most of their returns in the companies that turn that research into products, where capital needs are smaller and the customer relationship carries the value, the companies our mandate covers.

For founders

Building an AI company with heavy compute needs and a clear product? 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.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.