General Compute has closed a $100 million Series A round, valuing the inference-focused cloud provider at $800 million — a sharp jump from the $60 million valuation it carried after a seed round in May. The round was led by Blackbird Ventures and Abstract Ventures, and it follows a customer agreement worth $100 million in annual recurring revenue.
Unlike providers that chase model training, General Compute is built around inference: the everyday work of running trained models for users. That is the workload behind popular assistants such as ChatGPT, Claude and DeepSeek, where speed and cost per request matter more than raw training power.
The company buys specialised inference chips from Nvidia rivals, including a $400 million deal for SambaNova hardware that it describes as the first debt financing of inference chips. It has also added Positron and Cerebras as suppliers. Rather than building large data centres from scratch, it rents space and converts former bitcoin-mining sites, and it expects to deploy 75 megawatts of capacity in 2027, supplying bare-metal servers to fast inference cloud providers.
The timing is striking. Shares in rival neocloud CoreWeave have hit a 12-month low, and sceptics have warned that many such providers could fail. Chief executive Finn Puklowski pushed back, saying demand for compute keeps arriving weekly and that fears of a collapse are overstated. Investors in the round argued that chipmakers need partners able to scale in the physical world, while the real test remains whether inference speed can be sold profitably.
For a head-to-head look at two of the leading assistants running on this kind of infrastructure, see our ChatGPT vs Claude comparison.
Source: The AI Insider
