General Compute Launches ASIC-First Inference Cloud for Autonomous AI Agents
General Compute today announced its inference cloud platform built for AI agents, working with early partners now ahead of general availability on May 15, 2026. The platform runs on purpose-built AI accelerators rather than general-purpose GPUs. More information is available at generalcompute.com.
General Compute AI Inference Platform
SAN FRANCISCO — April 18, 2026 — General Compute Inc. today announced its inference cloud platform, which is designed for AI agent workloads. The company is working with early partners now, with general availability scheduled for May 15, 2026.
The platform runs on purpose-built AI accelerators rather than general-purpose graphics processors. Its architecture separates the prefill and decode stages of inference processing, allowing each stage to be scaled independently based on workload. The platform is built to serve AI agents that make high volumes of LLM inference and tool calls, including AI agents that provision their own compute programmatically.
“The last 20 years we built for developers, the next 20 we will build for agents. On General Compute, AI agents can sign up on their own and provision their own inference. Our docs and API are optimized for both human and AI agent consumption,” said Jason Goodison, co-founder and Chief Technology Officer of General Compute.
Platform Overview
The platform offers an industry-standard API, allowing developers to integrate it into existing applications with minimal code changes. AI agents and developers alike can sign up, provision API keys, and begin making inference calls programmatically.
At launch, the platform will offer access to a range of open-source LLMs across multiple model families and parameter sizes. Customers can also deploy their own models on the company’s infrastructure.
Infrastructure
General Compute’s data center infrastructure operates on hydroelectric power. The company states that its accelerator hardware is air-cooled, and that its racks operate at lower power densities than comparable installations built on general-purpose processors.
The company publishes technical performance data for its platform on its website.
Availability
General Compute is working with early partners now, with general availability beginning May 15, 2026. Enterprise inquiries regarding dedicated infrastructure, service level agreements, and capacity planning may be directed to jason@generalcompute.com
About General Compute
General Compute Inc. is an inference cloud company headquartered in California. The company was founded by Jason Goodison and Finn Puklowski.
Contact
Jason Goodison, Co-founder and Chief Technology Officer General Compute Inc. jason@generalcompute.com generalcompute.com
Frequently Asked Questions (FAQs)
1. What is the General Compute inference cloud platform?
The General Compute inference cloud is a platform designed specifically to run AI model inference workloads, especially for autonomous AI agents. It provides infrastructure where developers and agents process large volumes of requests efficiently. Instead of relying on general-purpose hardware, it uses purpose-built accelerators to improve performance.
2. How is this platform different from traditional GPU-based clouds?
Unlike traditional GPU-based clouds, this platform uses ASIC-based accelerators. As a result, it focuses specifically on optimizing inference tasks. In addition, it separates the prefill and decode stages, and this separation allows each stage to scale independently based on workload needs.
3. What are AI agents in this context?
AI agents are software systems that perform tasks, make decisions, and interact with tools using large language models. Moreover, these agents generate high volumes of inference requests. In many cases, they also operate programmatically without continuous human input.
4. Can developers integrate the platform into existing applications?
Yes, developers can integrate the platform easily because it provides an industry-standard API. Therefore, they connect their applications with minimal code changes. However, the level of adjustment still depends on the existing infrastructure and the specific use case.
5. What types of models are supported on the platform?
At launch, the platform supports various open-source large language models across different sizes and families. In addition, users can deploy their own models on the infrastructure, provided they meet compatibility and configuration requirements.
6. How do users get started with the platform?
Users start by signing up and generating API keys. Then, they begin making inference calls programmatically. Furthermore, both developers and AI agents interact with the system through APIs and documentation designed for both machine and human use.
7. What is meant by separating prefill and decode stages?
Inference processing includes two main stages: prefill and decode. By separating these stages, the system allows each one to scale independently. Consequently, this improves efficiency depending on workload type and request volume.
8. What infrastructure does General Compute use?
The company states that its infrastructure runs on hydroelectric power. It also uses air-cooled accelerator hardware. Additionally, rack power density differs from systems that rely on general-purpose processors, which improves energy and performance management.
9. When will the platform be generally available?
General availability is scheduled for May 15, 2026. Currently, early partners already use the platform. However, broader access will depend on rollout plans and system capacity.
10. Does the platform support enterprise-level requirements?
Yes, the company indicates support for enterprise-level requirements. For example, dedicated infrastructure and service level agreements may be available. However, final terms, capacity, and configurations depend on specific discussions and project needs.
Company Details
Organization: General Compute Inc
Contact Person: Jason Goodison
Website: https://generalcompute.com
Email: jason@generalcompute.com
Contact Number: +14257537667
Address: 440 North Barranca Avenue
Address 2: 3780
City: Covina
State: California
Country: United States
Release Id: 18042644077