Fireworks AI acquires Hathora to enhance AI reasoning capabilities

Fireworks AI Inc., an emerging company in the field of artificial intelligence inference, announced the acquisition of Hathora Inc. Hathora is a provider of real-time computing and server orchestration platforms focused on multiplayer gaming and AI inference. Fireworks AI CEO and co-founder Lin Qiao stated that the acquisition aims to acquire technological infrastructure and talent, rather than gaming clients. Through this acquisition, Fireworks AI plans to enhance its existing global computing orchestration capabilities.

Qiao explained the importance of this by linking the significance of gaming infrastructure to the development stage of AI infrastructure. Gamers may tolerate a decline in graphics quality or frame rate, but they can immediately notice interruptions in gameplay caused by latency. “AI inference also requires this millisecond-level decision-making,” Qiao emphasized.

Since its launch in 2022, Hathora officially released its platform at the 2023 Game Developers Conference. Recently, the company has expanded into the AI real-time workload domain and entered the speech model market.

With rapid industry changes, the speed of AI model delivery, especially inference and training speeds, is accelerating further. Fireworks will leverage this acquisition to accelerate the development of its global computing platform and aims to promote advancements in global orchestration, instant recovery, auto-scaling, low latency, and high-performance configurations.

Acquiring Hathora exemplifies how Fireworks’ strategic direction is evolving in inference and training. Fireworks aims not only to become an inference service provider but also to offer customized business solutions through pre-trained models, post-training functionalities, and automation layers.

Qiao stated that their goal is not just inference but continuous model customization. “Inference is just the result of all these setups we implement,” he added. Fireworks hopes to become a game-changer in the market for ongoing model fine-tuning and inference services.