General Compute raises $400M loan backed by inference chips
Original: Why the first GPU financiers are turning to inference chips in a $400 million deal
Why This Matters
Inference chip financing signals a structural market shift from model training toward cost-efficient AI deployment infrastructure.
AI inference cloud startup General Compute secured a $400 million loan from Upper90, backed by SambaNova inference chips as collateral — potentially the first such deal of its kind. The company aims to deliver 16x faster inference than GPU-based clouds using power-efficient SN50 chips.
General Compute, an AI inference cloud startup founded by CEO Finn Puklowski, has closed a $400 million loan from Upper90, a tech investment firm led by former Goldman Sachs quantitative trader Billy Libby. The deal may be the first to use inference-specific chips — rather than training GPUs — as collateral. General Compute had previously raised a $15 million seed round in May 2026 to build an inference neocloud around SambaNova's SN50 chips, which are designed for running already-trained AI models efficiently. The SN50 chips are power-efficient, require no water cooling, and can be deployed faster across a broader range of data centers. General Compute claims its infrastructure delivers 16 times faster inference than GPU-based clouds. Upper90's Libby pioneered chip-backed financing in 2021 when he financed GPU purchases for Crusoe, a deal he says was the first loan against advanced chip value. That model was later popularized by CoreWeave's IPO. Now, with GPUs considered mature and potentially over-bought, Upper90 is betting on inference-focused infrastructure. 'Everyone doesn't need a supercomputer, but they do need inference and AI,' Libby told TechCrunch. The move aligns with growing investor interest in open-source model infrastructure, including OpenRouter and Fireworks raising large rounds.