Fireworks AI, an artificial intelligence inference startup, has acquired Hathora Inc., a real-time compute and server orchestration platform. The acquisition aims to bolster Fireworks' global compute orchestration layer for AI inference and training.

Fireworks AI CEO Lin Qiao emphasized that the company sought Hathora for its talent and infrastructure, drawing parallels between the gaming industry's demand for low latency and the critical needs of AI inference. "That discipline, the obsession with every millisecond and every routing decision, is exactly what AI inference needs," Qiao stated.

Hathora, launched in 2023, built a container orchestration platform across 14 regions, supporting live gaming titles and expanding into real-time AI workloads. The acquisition is expected to accelerate Fireworks' development of a global computation platform requiring low-latency, high-performance orchestration, disaster recovery, and auto-scaling.

This move aligns with Fireworks' vision for agentic AI, where multimodality and rapid agent-to-agent interactions will be paramount. The company believes the future lies not in universal models, but in "millions of models," continuously customized for specific use cases. Fireworks aims to lead in continuous fine-tuning and low-latency inference at scale, positioning inference as the output of its automated customization process.