General Compute secures $100 million to expand inference cloud
General Compute has raised a $100 million Series A led by Blackbird Ventures and Abstract Ventures, taking the AI infrastructure startup to an $800 million post-money valuation. The round follows a May seed financing that valued the company at $60 million and comes after General Compute signed a large customer contract worth $100 million in annual recurring revenue.
Co-founded and led by CEO Finn Puklowski, the company is building a neocloud focused on AI inference rather than the GPU-heavy training infrastructure that has dominated much of the sector. General Compute supplies bare-metal servers to cloud providers that offer developers high-speed inference services.
An inference-first hardware strategy
General Compute has assembled a portfolio of specialised inference-chip suppliers: SambaNova, Positron and Cerebras. Its $400 million agreement to buy new SambaNova chips was described as the first debt financing of inference chips. The company’s supplier approach builds on General Compute’s SambaNova inference-chip strategy while extending its infrastructure portfolio to Positron and Cerebras.
The strategy places General Compute in a different segment from operators racing to deploy large fleets of GPUs. Puklowski said demand for compute remains strong, citing regular requests for access to TPUs, GPUs and other chips. He also said Anthropic’s S-1 had helped settle current market concerns.
Blackbird principal James Palmer, who co-led the financing, pointed to the demand for open-weight inference and for companies to own their own intelligence. Blackbird’s investment in inference cloud provider Baseten gave the firm a close view of that company’s growth over the preceding 10 months, Palmer said.
Power availability and financing remain central
Rather than construct massive data centres from the ground up, General Compute rents capacity in other companies’ facilities, uses smaller power sources and converts former bitcoin-mining sites to run AI models. The model is intended to ensure that chips can be deployed when they become available, without the same construction constraints faced by new data-centre projects.
Puklowski expects General Compute to deploy 75 megawatts of computing capacity in 2027 and said the business could have access to 200 megawatts of power. He characterised that potential supply as more power than the company could probably fund and equip with hardware.
That qualification underlines the company’s remaining challenge. Puklowski said capital raising across debt and equity would be a full-time responsibility. SambaNova CEO and co-founder Rodrigo Liang framed the issue in commercial terms: the market is full of promises of speed, but the important question is whether AI cloud compute can be operated profitably.
What the funding signals for buyers
For organisations procuring AI inference capacity, the round highlights that viable supply can come from specialised-chip platforms and repurposed infrastructure, not only from the largest GPU clouds. Buyers should weigh hardware availability, power access, deployment location and the economics of the service alongside raw performance claims.