Fireworks AI Advances Fresh Funding on Inference Cloud Boom

Bloomberg first reported in late May that Fireworks AI was in talks to raise fresh funding at a $15 billion post-money valuation, with Index Ventures set to co-lead. CNBC's latest coverage updates that same round. Take the specifics as reported, not settled, but the trajectory lines up with what other outlets have been tracking for months.

The relevant baseline is short. Fireworks closed a $250 million Series C in October 2025 at a $4 billion post-money valuation. Sacra estimated the company hit roughly $800 million in annualized revenue by May 2026, up from about $305 million at the end of 2025. If either of those anchors is close to right, the multi-billion re-rating priced into this round is being underwritten by real inference volume rather than pure narrative.

Founded by Lin Qiao, a former Meta engineering lead who co-created PyTorch, Fireworks sits in a narrowing set of independent inference clouds, the layer that runs open-source and fine-tuned models on GPU capacity for developers who do not want to build directly on hyperscaler stacks. Its customer roster already includes Cursor, Perplexity, Notion, Uber, Shopify, and Sourcegraph. Nvidia and AMD are on the cap table, which matters because chip access is the single hardest input to secure in this business right now.

The honest caveat is that most of the freshest revenue and valuation numbers in this story are either estimates from third parties or figures reported by outlets before the round formally closed. What the reporting does not give you is the actual gross margin at which Fireworks is serving those tokens, or how much of the revenue is durable enterprise commitment versus spikier developer usage.

The forward question is whether the independent inference tier can hold enough of the market once hyperscalers make hosted open-model serving a first-class product. If the answer is yes, cap tables like this one look cheap. If not, the reset will be fast.