Tech news in 3 minutes
Why the first GPU financiers are turning to inference chips in a $400 million deal
**General Compute secures $400M loan from Upper90 using inference-specific chips as collateral**, a deal that marks a first for the emerging AI inference cloud sector and signals growing financial market appetite for cost-efficient, Nvidia-alternative infrastructure. The startup, founded by CEO Finn Puklowski and CTO Jason Goodison, raised a $15 million seed round in May to build a neocloud powered by SambaNova’s SN50 chips—purpose-built for inference, not training. These chips are power-efficient, don’t require water cooling, and can be deployed faster across more data centers. General Compute claims they deliver 16x faster inference than GPU-based clouds. Upper90, a tech investment firm led by former Goldman Sachs quant Billy Libby, structured the loan—possibly the first to put inference-specific chips as collateral. Libby previously financed GPU purchases for Crusoe in 2021, pioneering chip-backed loans before CoreWeave turned the model into an IPO blockbuster. Now, with GPUs well-understood and possibly over-bought, Upper90 is betting on the next wave: open-source inference. “Everyone doesn’t need a supercomputer, but they do need inference and AI,” Libby said. The financing reflects a broader market shift: open-source model providers like OpenRouter and Fireworks are raising large rounds, new chipmakers like Groq and Cerebras attract acquirer interest, and TensorWave makes a similar bet on AMD. As alternative chips scale, General Compute’s access to non-Nvidia silicon gives it a cost-efficiency advantage. “This is the first signal of capital organizing itself and the fragmenting of Nvidia’s monopolistic dominance,” Puklowski said.