All news

Frontier labs now trade equity for cloud capacity

A graphite server module and a white marble block bound together with cobalt cord on a pale table

Microsoft's multibillion-dollar bet on OpenAI ties every workload of the lab to Azure. Capital and compute now arrive as a single package.

Microsoft's new investment in OpenAI has no published price, but it has a clear condition: Azure stays the lab's exclusive cloud provider for all workloads across research, products and API services. The companies announced this third phase of their partnership on 23 January, a multiyear, multibillion-dollar investment that follows Microsoft's earlier investments in 2019, when it put in $1 billion, and in 2021.

That condition sets the financing template for a frontier lab. The lead investor is also the cloud, so money and compute arrive as one package and every dollar the lab spends on compute is spent with its investor. The value of an independent lab now depends on its access to compute as much as on its revenue.

Both sides get what a plain equity cheque cannot buy. Microsoft gains guaranteed demand for its data centres and a product edge across its software: it will build more specialised supercomputing systems for OpenAI's research and deploy OpenAI's models across its consumer and enterprise products. OpenAI gains capacity that no ordinary venture round could fund, even the large rounds labs raised last spring. On 3 February Anthropic selected Google Cloud to train and serve its models, and the frontier is concentrating in a few pairs of lab and cloud.

Revenue arrived on the same timetable. Nine days after the announcement OpenAI launched ChatGPT Plus at $20 a month, its first paid consumer plan, proof that consumers will pay for the assistant that reached a million users in five days.

For an AI-native application the arithmetic runs the other way. Models from these pairs will reach every company through APIs and cloud marketplaces, so a startup can use frontier capability without building a lab, yet the same models will ship inside office software, developer tools and operating systems owned by the clouds. The defensible ground is the workflow, the proprietary data and the customer relationship. Pricing has to keep the model cost per user in view, because the bill for every call goes to a single upstream supplier, and cloud credits become a commercial decision once capacity comes with a commitment to one provider.

The bundle still has to prove itself at both ends. How quickly Microsoft builds OpenAI models into its products, and how much of its cloud growth it attributes to AI, will show whether the cloud side of the trade pays. If other clouds sign similar pairings with independent labs, the open question is whether the terms carry equity, exclusivity or both. Demand is the last unknown: the uptake of ChatGPT Plus, and of the lower-cost and business plans OpenAI says it is exploring, will show what users pay for direct access to a model.

The frontier model layer is becoming a strategic asset of the largest clouds, and financial investors will find their returns one level up, in the companies that turn those models into products for specific industries. That is where Fanaura looks: AI-native companies that own the task and the customer and treat the model as a supplier they can change.

For founders

Building an AI-native product for a specific industry on top of frontier models? 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.