AI infrastructure company Fireworks AI has closed a $1.505 billion Series D at a $17.5 billion valuation, the largest financing yet for a company selling specialized inference to enterprises. The round announced on 15 July was led by Atreides Management, Index Ventures and TCV, with Evantic Capital, Lightspeed Venture Partners, NVIDIA, 20VC, Bessemer Venture Partners and Menlo Ventures also participating.
From general LLMs to specialized intelligence
Fireworks says more than 95 percent of the more than 40 trillion tokens served on its platform every day now come from models specialized on customers' own data, rather than general-purpose frontier models. That pipeline of fine-tuned deployments has taken the company to more than $1 billion in annualized revenue run rate, up roughly 5x year-on-year, with daily token volume rising from 15 trillion to over 40 trillion over the same period.
Enterprise flywheel and open weights
The company's pitch to CFOs is that as open-weight models close in on closed frontier systems, enterprises no longer need to rent generic intelligence; instead they can fine-tune open models on proprietary data and deploy them via Fireworks at a fraction of the cost. Customers cited by the company include Cursor's coding models and Harvey's legal AI, both of which run on the Fireworks stack.
Capital for compute and go-to-market
The proceeds will fund a further expansion of Fireworks' compute footprint and deepen partnerships with hyperscalers, including Microsoft and NVIDIA, and its engineering team. The valuation lands as investors continue to plough record capital into the AI infrastructure layer, alongside recent mega-rounds like Moonshot AI's 2.8-trillion-parameter Kimi K3 release and Google DeepMind's Gemini 3.5 Pro launch. It also comes days after Emergent's $130M Series C highlighted an AI app factory unicorn emerging from Bangalore.
Reporting based on coverage from Fireworks AI, BusinessWire and SiliconANGLE.