General Compute Secures Up to $400 Million in Debt to Scale the World's Fastest Inference Neocloud

General Compute, building the world's fastest AI cloud, led by Finn Puklowski and Jason Goodison, has announced a committed debt facility of up to $400 million from Upper90 Capital Management, LP (“Upper90”). The financing begins at $100 million and scales with customer demand, allowing General Compute to quickly build one of the world’s largest inference neoclouds to deliver AI results 16x faster than current GPUs. 

$400M Debt Facility to Expand AI Compute Infrastructure.

General Compute has secured a $400 million debt facility to accelerate the deployment of next-generation AI compute infrastructure. As one of the largest financings for an ASIC-based cloud provider, the investment supports the company's strategy of building on independent AI inference silicon rather than relying exclusively on GPUs.

Backed by more than $300 million in secured, price-protected chip supply, General Compute aims to scale ASIC-powered inference infrastructure to deliver greater performance and energy efficiency. The financing, led by Upper90, also an equity investor, aligns long-term capital with the company’s growth as demand for AI inference continues to increase.

Upper90 highlighted General Compute’s differentiated approach to AI infrastructure, noting that its partnership with SambaNova provides access to inference silicon that delivers significantly higher power efficiency than traditional GPU-based systems while supporting lower-cost, high-performance AI compute.

The GPU Inference Challenge.

As global AI adoption accelerates, token consumption is expected to grow rapidly over the coming years. While GPU-based infrastructure has been effective for AI training, it remains costly and power-intensive for inference. The latest GPU systems require significantly higher power and specialized cooling, creating infrastructure constraints that slow the expansion of AI compute capacity.

General Compute's Solution.

General Compute is building an AI inference neocloud powered by specialized inference chips from SambaNova. Designed for high-speed, energy-efficient AI inference, the platform operates with significantly lower power requirements than traditional GPU infrastructure while eliminating the need for liquid cooling, enabling faster AI responses and more scalable deployment.

Key Features:

• Delivers significantly faster AI inference and response times than conventional GPU-based clouds.

• Provides higher throughput to support demanding AI workloads at scale.

• Uses air-cooled infrastructure, enabling faster deployment without specialized data center upgrades.

• Improves energy efficiency, reducing the cost of AI inference.

• Supports seamless compatibility with leading frontier AI models.

• Designed for agentic AI workflows with API and MCP compatibility.

“General Compute is focused on delivering high-performance AI inference for frontier models at speed. With an infrastructure strategy independent of a single chip supplier, the company can rapidly deploy next-generation inference silicon into existing data centers to meet growing demand for AI workloads.”

— Finn Puklowski, Co-founder and CEO, General Compute