General Compute

General Compute Inc. is a California-based company that provides an inference cloud platform specifically designed for high-volume, latency-sensitive AI agent workloads. The company was founded by Finn Puklowski and Jason Goodison, who serves as the Chief Technology Officer. General Compute positions itself as an "ASIC-first AI neocloud," aiming to address the performance constraints of traditional GPU-based clouds, which were primarily built for AI training rather than inference.

The core of General Compute's offering is a platform built on purpose-built AI accelerators (ASICs) instead of general-purpose GPUs. This specialized hardware architecture separates the prefill and decode stages of inference processing, allowing each stage to be scaled independently to match workload demands. The company states this approach delivers significantly faster token generation, which is critical for applications like coding assistants, voice AI, and autonomous systems that rely on rapid, repeated calls to large language models (LLMs). The platform is designed to be accessible to both human developers and AI agents, offering an industry-standard API that enables agents to programmatically sign up, provision API keys, and run inference calls without human intervention.

The business model appears to be a form of cloud computing service, likely usage-based, where customers can run a range of open-source LLMs or deploy their own custom models on General Compute's infrastructure. This targets a growing market of enterprises deploying AI agents for complex workflows in software development, finance, and customer operations. The company announced it was working with early partners ahead of a general availability date of May 15, 2026. As part of its infrastructure, the company notes that its data centers operate on hydroelectric power and utilize air-cooled hardware.

Keywords: AI inference cloud, ASIC accelerators, high-speed inference, large language models, AI agents, token generation, custom silicon, cloud platform, AI infrastructure, agentic AI, low-latency AI, autonomous systems, inference optimization, purpose-built hardware, API-driven compute, cloud computing services, neocloud, decode-prefill separation, AI workloads, Jason Goodison, Finn Puklowski