General Compute raises $100 million to build an AI cloud using Nvidia rivals’ chips

General Compute has fresh equity to build an AI cloud around chips from Nvidia rivals, with its first major deployments planned for early 2027. The company announced a $100 million Series A on October 27, co-led by Blackbird and Abstract. TechCrunch reported that the round values the business at $800 million after the investment.

Fuse Venture Partners and Village Global also participated. The equity round builds on a previously announced $400 million debt facility from Upper90, giving General Compute two sources of financing for its infrastructure plans: investor capital and borrowing.

A large customer agreement helped make the financing possible. TechCrunch reported that the deal is worth $100 million in annual recurring revenue. That is the reported value of one customer agreement, not a disclosed total for the company’s current revenue. General Compute’s valuation was $60 million after a seed round in May.

A route to customers for specialized chips

General Compute focuses on inference: running AI models to produce responses. It says it finances, deploys and operates graphics processing units, or GPUs, alongside purpose-built inference chips from Cerebras, SambaNova and Positron AI. Its stated approach is to match each workload to the hardware that runs it best, rather than offer only one kind of chip.

Its customers include cloud companies that offer developers high-speed inference. General Compute supplies the bare metal—raw servers without a software layer—on which those customers build their services. The company describes two ways to buy access without having to procure or manage the underlying infrastructure:

• Dedicated bare-metal infrastructure, giving customers servers on which to run their own offerings.
• A service-level agreement guaranteeing tokens per second, a measure of how quickly the system produces AI output.

Blackbird principal James Palmer told TechCrunch that Cerebras, Positron and SambaNova need distribution partners able to handle the physical work of scaling their chips. He also cited Blackbird’s experience investing in inference-cloud company Baseten as part of the firm’s confidence in General Compute. That is an investor’s rationale for the round, rather than a measure of General Compute’s operating performance.

Rented space instead of a ground-up buildout

General Compute is not pursuing its expansion by building huge data centers from scratch. It rents space in other operators’ facilities, uses smaller power sources and converts former bitcoin-mining sites into locations for running AI models. The approach sidesteps the ground-up construction work that would otherwise accompany a large new facility.

The company’s announcement puts its first major deployments in the first quarter of 2027, with more than 50 megawatts of capacity in its pipeline. In his interview with TechCrunch, co-founder and CEO Finn Puklowski gave a broader full-year expectation: deploying 75 megawatts during 2027. Both figures describe future capacity, not computing infrastructure already operating.

capital is going to be a full-time job for me, both on debt and equity

Finn Puklowski, General Compute co-founder and CEO, speaking to TechCrunch

Power access is not the same as funded hardware

Puklowski expects access to power could exceed what the company can equip. He told TechCrunch he suspected General Compute would have access to 200 megawatts—more than it could probably fund and obtain hardware for. His estimate separates finding somewhere to plug in chips from paying for and securing the chips themselves.

SambaNova CEO and co-founder Rodrigo Liang, whose company supplies General Compute, put the commercial test plainly to TechCrunch: everyone is selling speed; the question is whether they can make money doing it. His comment places profitability alongside fast output as a test of the business, rather than treating speed alone as the result that matters.