RadixArk launches with $100M seed to build open AI inference infrastructure
The AMW Read
RadixArk is a new entrant to the AI infrastructure segment, with an unusually large seed round ($100M) that updates the player map and signals capital cycle dynamics. The amount meets the cross.§D threshold.
RadixArk launches with $100M seed to build open AI inference infrastructure
RadixArk, a Palo Alto-based AI infrastructure startup founded by Ying Sheng and Banghua Zhu, launched today with $100 million in seed funding. The round was led by Accel and Spark Capital, with participation from AMD, MediaTek, NVentures, Salience Capital, and prominent angels including John Schulman and Thomas Wolf. The company plans to use the capital to grow the open-source inference engine SGLang, support emerging AI model architectures and hardware platforms, and build large-scale inference and training infrastructure.
Why it matters: RadixArk represents a capital-intensive bet that open-source inference infrastructure can become a foundational layer for the AI industry. The $100M seed — among the largest ever — signals that investors see the inference stack as a critical bottleneck as enterprises deploy frontier models at scale. The company dual-tracked on two open-source projects (SGLang for inference, Miles for RL training) positions it to compete with established players like vLLM while also addressing reinforcement learning workloads, a growing need for post-training optimization.
Grounded take: This launch fits the pattern of infrastructure-first companies raising massive rounds to standardize and commoditize the model-serving layer. RadixArk's open-source DNA and co-founder track record (Sheng contributed to SGLang; Zhu has RL research background) give it credibility, but the market is crowded with well-funded alternatives. The participation of AMD and MediaTek suggests RadixArk aims to optimize for diverse hardware, potentially breaking NVIDIA's lock-in. The real test will be enterprise adoption and whether open-source infrastructure can generate revenue beyond consulting and managed services.
