Faderless AI raises $20 million… Expanding open-source AI inference infrastructure
Boundless AI secures $20 million investment… Expanding open-source AI inference infrastructure

Open-source artificial intelligence (AI) model hosting startup Federless AI has secured a new funding round of $20 million, equivalent to approximately 29.38B Korean won. The company plans to use this capital as a foundation to expand its global infrastructure and launch a brand-new professional open model trading marketplace.

Founded in 2024, Federless AI claims to have built a “neutral hosting layer” that supports enterprises in running open-source AI without being tied to specific clouds or hardware. Currently, the company offers over 30k open models including language, vision, and audio, and is described as one of the fastest-growing inference partners within Hugging Face.

The company’s core is a single application programming interface (API). Customers can access multiple models through just one API, while the actual graphics processing unit (GPU) allocation, scaling, and traffic distribution are managed by Federless AI. It explains that for enterprises, this reduces the barrier to adopting experimental open-source AI by eliminating the need to build infrastructure for each individual model.

Federless AI stated in a release: “We have not only built a ‘inference engine’ but also created an ‘AI optimization stack’ that integrates inference, models, and workflow optimization.” It emphasized, “Through this architecture, we can achieve performance and cost-efficiency levels even on a closed platform that are difficult to reach with a single model, at a scale of 30k models.”

RWKV developer joins… Led by AMD and Airbus Ventures

Co-founder Eugene Cheah (CEO) is one of the creators of the open-source model architecture “RWKV,” developed under the Linux Foundation. RWKV features a recurrent structure as an alternative to existing Transformer-based systems. The company has assembled a core team including CTO Harrison Vanderbyl and COO Wesley George, primarily composed of members from the RWKV open-source community.

This Series A funding round was co-led by AMD Ventures and Airbus Ventures. BMW i Ventures, Kickstart Ventures, Panoia Ventures, and WaveMaker Ventures also participated. Previously, Federless AI had received support from Airbus Ventures in 2025, completing a seed round of $5 million.

The new funds will be focused on three areas. First, expanding infrastructure to more regions; second, launching a dedicated marketplace for trading fine-tuned models and open models optimized for specific purposes; and third, strengthening integration with various hardware architectures, including AMD accelerators, to further reduce inference costs.

Enterprises evaluate open models as alternatives to closed AI

This funding also aligns with trends in the enterprise AI market. Recently, corporate clients have been paying attention to the performance of closed, cutting-edge models like OpenAI and Anthropic, while also weighing the high costs and platform dependency risks. Meanwhile, companies such as Meta Platforms, Mistral AI, and Alibaba Group Holdings have provided alternatives by increasing the release of open-weight models.

In this environment, open-source AI infrastructure companies like Federless AI can target both “broadening choices” and “reducing costs.” Especially as enterprises increasingly compare various models or rapidly test industry-specific models, the utilization of inference platforms based on a single API may further rise.

Ultimately, this round of investment is not just a simple capital raise but also indicates that the open-source AI ecosystem has officially entered the infrastructure competition stage within the enterprise market. Whether Federless AI can attract actual demand in the future will be a key variable in measuring the commercialization speed of open models.

TP AI Note: This article was summarized using TokenPost.ai’s basic language model. The main content may be omitted or may differ from the actual details.