Featherless.ai, a serverless inference platform founded in 2023 by Eugene Cheah, Harrison Vanderbyl, and Wesley George, has secured $20 million in Series A funding co-led by AMD Ventures and Airbus Ventures.
Additional investors include BMW i Ventures, Kickstart Ventures, Panache Ventures, and Wavemaker Ventures.
Featherless addresses structural challenges that are often overlooked outside the developer community. Hugging Face hosts over 30,000 openweight AI models, many of which are tailored to specific languages, domains, and tasks that OpenAI and Anthropic’s flagship models don’t handle well, and accessing these models in production remains difficult.
“Typically, the only models available from providers are the most popular ones. Models trained in more niche fields are very difficult to access. Making them available online and continuously at a price that doesn’t require you to rent thousands of dollars of computing to talk to a chatbot that speaks your language is the genesis of Featherless,” says George. Technology funding news.
AI, Replicate, and Baseten work together to provide API access to open models, and Groq provides fast inference on custom silicon. Featherless argues that none are truly neutral, as each has hardware preferences, model partnerships, or ecosystem adjustments that reproduce the lock-in that the platform seeks to avoid.
The startup stands out with its hot-swap technology, which loads models into GPU memory on demand within 5 seconds and releases them when idle. This enables a flat-rate pricing model that provides fixed monthly capacity rather than billing per token.
“Most inference providers have 50 to 100 models available in the public cloud. We have an entire catalog of 30,000 models available online. You can’t run 30,000 models by dedicating $2,000 of hardware to each model. That’s what our competitors do. That’s differentiation,” explains George.
Businesses gain cost predictability not available with large cloud providers. This makes implementation economically viable for the first time for niche bespoke models.
The lineup of investors reflects the company’s global-first attitude. AMD’s involvement is strategically unique. Featherless ensures the most popular open models run natively on AMD’s ROCm platform, providing a competitive alternative to Nvidia’s ecosystem.
“AMD knows we can do great things with their hardware, and they’re very committed to open source. It’s a very natural fit,” George says. Airbus, which supported the company in the seed stage, is focused on implementing the open weight model in enterprise environments.
This sovereignty angle is increasingly becoming the company’s biggest growth opportunity.
“A year ago, there were still questions about whether open models had enough intelligence to do productive work. But that’s not the case anymore. The focus has now shifted to who controls the AI. Especially in markets outside the US, there’s a big move to control the models, the infrastructure, and the freedom to take what you build wherever you want,” George concludes.
The funding will support infrastructure expansion into new geographies, development of a specialized open model marketplace, and enhanced integration with Nvidia’s cross-platform hardware architecture.
