Inferact

Inferact is an AI infrastructure company founded by the creators and core maintainers of vLLM, an open-source large language model inference engine. Its stated mission is to grow vLLM into the world's AI inference engine and to make inference cheaper and faster. The company frames inference as an unsolved and worsening problem: models are growing larger, architectures are proliferating (mixture-of-experts, multimodal, agentic), and hardware is fragmenting across more accelerators and programming models, widening the gap between model capability and the systems that serve them. Inferact argues inference is shifting from a fraction of compute to the majority through test-time compute, reinforcement-learning training loops and synthetic data generation.

The company positions itself at the intersection of model innovation and hardware diversity, working with model vendors for day-zero support of new architectures and with hardware vendors integrating new silicon. Its long-term goal is to make deploying a frontier model at scale as simple as spinning up a serverless database, absorbing infrastructure complexity into its platform rather than requiring dedicated infrastructure teams. Sequoia Capital describes Inferact as supporting 200+ model architectures across major accelerator platforms with day-zero compatibility, while the company's own site cites vLLM support for 500+ model architectures and 200+ accelerator types.

Inferact was founded in 2025 and is based in San Francisco, California. Named founding members include Simon Mo (CEO), Woosuk Kwon, Kaichao You, Roger Wang, Joseph Gonzalez and Ion Stoica. In January 2026 it announced a $150 million seed round at an $800 million valuation.

Founding story

The founding team consists of the creators and core maintainers of vLLM, the open-source LLM inference engine incubated in 2023 at Ion Stoica's UC Berkeley lab. Founding members listed on the company site are Simon Mo, Woosuk Kwon, Kaichao You, Roger Wang, Joseph Gonzalez and Ion Stoica. The team says it has maintained the engine since its first commit and deployed it at frontier scale in both research and production, and started Inferact to build a universal inference layer that can run any model on any chip.

Business model

Inferact commercializes the open-source vLLM inference engine, building enterprise inference systems and a platform for serving models across hardware while contributing optimizations back to the open-source project. Sources state funding will be used to commercialize vLLM, scale the platform for enterprise AI inference, and expand adoption among existing and new customers; specific pricing or contract structures are not disclosed in the sources.

Not disclosed in the sources beyond an intent to commercialize vLLM and scale a platform for enterprise AI inference.

Traction

Traction is primarily ecosystem-based: vLLM is described as the most popular open-source LLM inference engine, with 2,000+ contributors, 500+ supported model architectures and 200+ supported accelerator types, and is said to power inference at global scale for frontier labs, hyperscalers and startups. Reported users of vLLM include Amazon's cloud service and shopping app. Investor validation includes a $150 million seed at an $800 million valuation.

Latest developments

In January 2026 Inferact announced its $150 million seed round at an $800 million valuation, co-led by Andreessen Horowitz and Lightspeed Venture Partners with Sequoia Capital, Altimeter Capital, Redpoint Ventures and ZhenFund participating, and appeared on the a16z podcast to describe its universal inference layer vision. A company news entry dated August 12, 2026 announces full certification of Kimi K3 vendor verification and the availability of its production inference stack.