Fireworks AI

Former Meta engineers Lin Qiao and David Ha established Fireworks AI in San Mateo, California, during 2022 to tackle the specific bottlenecks of running open-source large language models in production. The company built its reputation on solving the messy reality of deploying generative artificial intelligence at scale rather than offering raw compute or building proprietary closed models. Their platform targets businesses needing to integrate advanced AI capabilities without managing complex backend systems, optimizing how massive models process data to reduce latency and costs while maintaining high reliability.

The firm operates a globally distributed cloud infrastructure designed to handle internet-scale workloads, processing over 13 trillion tokens as of mid-2026. In March 2026, they expanded their market reach by integrating their technology into Microsoft Foundry, bringing high-performance, low-latency open model inference directly to the Azure ecosystem. This strategic move positioned them as a critical partner for deploying AI on cloud infrastructure alongside major hardware players like Nvidia.

By July 2026, the company achieved decacorn status after closing a $1.5 billion Series D funding round at a $17.5 billion valuation. Investors including Nvidia backed this latest round as the company pivoted toward specialized, smaller models and efficient inference rather than competing directly with general-purpose foundation model creators. Recent developments highlight a shift toward supporting AI agents and post-training techniques, allowing startups to customize models and own their intelligence through specialized tuning. [edit]