Featherless Raises $20M Series A to Power Open-Source AI Infrastructure

Featherless AI, enabling serverless inference via GPU orchestration and a model load-balancing system, led by Eugene Cheah and the team, has raised a $20 million Series A financing round co-led by AMD Ventures and Airbus Ventures, with participation from BMW i Ventures, Kickstart Ventures, Panache Ventures, and Wavemaker Ventures.

AI is entering a new phase focused less on capability and more on control, how models are run, where they are deployed, and who governs them.

Today, AI power is concentrated across a few countries, companies, and compute providers. Even as the ecosystem is described as open, access, pricing, and influence remain shaped by a small number of key bottlenecks.

Featherless was built to broaden access.

A neutral layer for open models, independent of hyperscalers, chipmakers, or proprietary ecosystems. Its infrastructure spans the EU and the US, with a founding team distributed across Canada, Singapore, and Australia, and teams based in San Francisco, Toronto, Singapore, and across Europe.

The Fastest-Growing Open Model Platform

Featherless is a rapidly growing inference partner in the Hugging Face ecosystem, serving 30,000+ open models across language, vision, audio, and multimodal use cases, enabling instant access without infrastructure overhead.

To operate at this scale, an integrated AI optimization stack has been built, combining inference, model, and workflow optimization to improve performance and cost efficiency across diverse workloads.

The system is informed by deep research, including contributions to RWKV and ongoing work through Recursal Labs, advancing open model development at the foundation level.

Open Hardware, Open Stack

True AI independence requires hardware diversity. Through a strategic partnership with AMD, leading open-source models are supported natively on AMD’s open-source ROCm platform.

The Home for Open-Source AI Apps

The next wave of AI is an ecosystem of applications: agentic workflows that reason, plan, and act; personal assistants owned by users; and tools built by independent developers.

Today, most applications still funnel builders into closed model ecosystems. Featherless breaks that dependency.

An open-source agent runtime has been built as the foundation for a new generation of AI applications powered entirely by open models, without reliance on closed APIs. This enables full control over models, data, and deployment.

The resulting ecosystem of tools, agents, and workflows is designed to be community-driven and broadly accessible.

What This Funding Enables

With $20M in funding, expansion is focused on global reach and deeper technical integration:

1) Models

Expanding the open model library for instant usability and faster access for creators.

2) Agents

Shipping an open-source agent runtime for the next generation of AI applications.

3) Optimization

Lowering inference costs through systems-level improvements across hardware, including AMD/ROCm support.

4) Enterprise

Scaling secure, production-grade deployments with private, sovereign environments for enterprise workloads.