General Compute's Series A was co-led by Blackbird and Abstract, TechCrunch reported. The company says it has a customer deal worth $100 million in annual recurring revenue.
Venture investors are still willing to pay sharply higher prices for companies that sell AI computing capacity.
General Compute, which provides computing power for running AI models, raised $100 million in a Series A round at an $800 million post-money valuation, TechCrunch reported on Tuesday. The round was co-led by Blackbird and Abstract. The company's seed round in May valued it at $60 million post-money.
The company cites a customer deal worth $100 million in annual recurring revenue.
The step-up
The valuation rose about 13 times in roughly five months. General Compute is pointing to a single large contract to explain why.
A single large contract is both the strongest argument for the price and its main risk. It gives the company revenue that most Series A startups do not have. It also makes the company dependent on one customer, whose name and contract terms have not been disclosed.
Where it fits
General Compute focuses on inference, the computing used to run AI models after they are trained, as opposed to training them. Demand for inference grows as more businesses put AI tools into daily use.
Two readings
One reading is that this is what a fast-growing infrastructure company looks like in this market: a large contract, real revenue and investors willing to pay for it early.
The other is that a 13-fold step-up on the strength of one company-described contract leaves little room for error. The customer is unnamed, the contract's terms are undisclosed, and a smaller cloud provider competes with much larger ones for both chips and customers.
What to watch
Disclosure of the customer or contract terms, the company's next funding round, and how it secures chips and data-center capacity to deliver on the contract.