Featherless.ai Raises $20M Series A Co-Led by AMD Ventures and Airbus Ventures to Scale Serverless Open-Model Inference (April 30, 2026)

Singapore-co-founded Featherless.ai on April 30, 2026 announced a $20 million Series A co-led by AMD Ventures and Airbus Ventures, with participation from BMW i Ventures, Kickstart Ventures, Panache Ventures, and Wavemaker Ventures. The serverless inference platform now hosts more than 30,000 open-source AI models and bills customers at flat monthly rates rather than per token, making it the largest open-model inference provider on Hugging Face.

What Happened

Featherless.ai disclosed the round in a blog post titled "Featherless Raises $20M Series A to Power Open-Source AI Infrastructure" on April 30, 2026. The funding closes a year-long stretch in which the startup grew from $5M seed (March 2025) to today's Series A, alongside a Hugging Face partnership that elevated it to that platform's largest LLM inference provider with more than 6,700 supported language models on launch and roughly 30,000 open models across language, vision, and audio today.

CEO and co-founder Eugene Cheah said the round will fund three priorities: scaling global serverless capacity, launching a dedicated marketplace for specialised open models, and deepening native support for AMD ROCm so popular open-weight models run on AMD Instinct GPUs without proprietary CUDA dependencies. The company hosts core infrastructure in the US and EU, with team members across Canada, Europe, the US, Singapore, and Australia.

Key Details

• $20M Series A, co-led by AMD Ventures and Airbus Ventures — strategic capital, not just financial. AMD's involvement underwrites the ROCm-native roadmap; Airbus Ventures' involvement is a deliberate sovereign-AI signal toward European customers.
• 30,000+ open models supported — including DeepSeek, Llama, Qwen, Mistral, and Google Gemma variants, plus vision and audio models from across Hugging Face.
• Flat-rate pricing starting at $10/month — Featherless replaces per-token billing with fixed monthly capacity. The platform's hot-swap technique loads models into GPU memory in under five seconds and unloads them when idle to make the economics work.
• Largest LLM inference provider on Hugging Face — a status confirmed by Hugging Face's own blog post on inference providers earlier in 2026.
• Total raised to date: $25M ($5M seed in March 2025 + $20M Series A April 2026).

What Developers and Users Are Saying

The Hacker News thread on Featherless' Series A trended in the top 30 within hours of the announcement, with developers focused on two questions: how the flat-rate model holds up under bursty workloads, and whether the 5-second hot-swap latency is acceptable for production chatbots. Several commenters running side projects praised the cost predictability — one said the $10/month tier replaced an $80/month OpenRouter bill on long-tail open-weight models. Skeptics pointed out that frontier closed models from OpenAI and Anthropic remain unavailable, which Featherless does not contest.

On X (Twitter), AMD's developer relations account amplified the news, framing Featherless as "the easiest way to run open-source models on AMD Instinct in production." European AI buyers reacted positively to Airbus Ventures' involvement, reading it as a sovereign-AI alignment signal at a time when the EU AI Act's compute-disclosure thresholds are starting to bite.

What This Means for Developers

If you are running open-source LLMs at small or hobby scale, Featherless is now a credible flat-rate alternative to OpenRouter, Replicate, and Together AI for the long tail of Hugging Face models. The economics tilt in your favor when you run many low-traffic models rather than a single high-throughput one. The AMD ROCm story matters for teams that want to avoid NVIDIA lock-in or that already run on AMD MI300 hardware.

The Hugging Face integration is the easiest entry point: pick Featherless from the Inference Endpoints dropdown on any model page and you are billed through Hugging Face. For direct access, the Featherless API uses the OpenAI-compatible chat-completions schema, so most existing client SDKs work without changes.

What's Next

Featherless' near-term roadmap focuses on the open-model marketplace launch, expanded ROCm coverage, and additional EU and Asia-Pacific regions. The company has not committed to a Series B timeline. Watch for further AMD-specific announcements at AMD Advancing AI 2026 and follow the official Featherless blog for product updates.

Sources

• Featherless Raises $20M Series A to Power Open-Source AI Infrastructure — primary announcement on the company blog.
• SiliconANGLE — Featherless.ai pulls in $20M to scale serverless hosting for open-source AI models
• Tech.eu — Featherless.ai secures $20M to expand serverless platform
• TNGlobal — Singapore co-founded Featherless.ai raises $20M Series A
• TFN — AMD and Airbus back Featherless.ai's $20M Series A
• Hugging Face Blog — Featherless AI on Hugging Face Inference Providers