Featherless bets open-source AI needs a neutral serverless layer, not walled gardens

Featherless.ai was co-founded in Singapore by the team behind RWKV, an open-source neural architecture designed to challenge transformers, with Eugene Cheah as CEO. Their read on the AI market: the first wave belonged to proprietary, closed ecosystems, but the next wave would be millions of specialised, fine-tuned open models — and those models would need infrastructure that could serve them reliably and cheaply without locking buyers into one cloud or chipmaker.

The company built a serverless inference platform that gives developers API access to more than 30,000 open models spanning language, vision and audio, claiming to be Hugging Face's fastest-growing inference partner. It runs on hardware-agnostic infrastructure, including a strategic collaboration making popular open models run natively on AMD's ROCm platform, and hosts its core infrastructure in the US and EU with a global team across Canada, Europe, the US, Singapore and Australia — a structure it sells as sovereign AI for customers with jurisdictional and privacy constraints.

On 2026-04-30 Featherless.ai announced a $20 million Series A co-led by AMD Ventures and Airbus Ventures, with BMW i Ventures, Kickstart Ventures, Panache Ventures and Wavemaker Ventures participating. AMD's venture arm framed the company as the foundation for a competitive AI ecosystem; Airbus Ventures argued the bottleneck was not model availability but serving the long tail reliably at scale, which Featherless addresses by loading models in seconds while keeping GPUs productively utilised.

As of May 2026 the company was using the capital to scale global infrastructure, launch a marketplace for specialised open models and deepen hardware integration, with no revenue or customer figures disclosed. The bet — that neutrality and hardware diversity, not frontier performance, win the open-model era — was still being tested against well-funded proprietary platforms.