Skip to content
NewsFunding & M&A

Fireworks AI raises $1.505 billion Series D at $17.5 billion valuation

· by Pondero Newsdesk

The short version

The open-weight inference cloud crossed $1 billion ARR and 40 trillion tokens per day before closing a round led by Atreides Management, Index Ventures, TCV, and Nvidia.

Fireworks AI raises $1.505 billion Series D at $17.5 billion valuation

Open-weight inference infrastructure cleared its largest funding milestone on record this week. Fireworks AI announced a $1.505 billion Series D on July 15, 2026, valuing the company at $17.5 billion, per its company blog. By the time the round closed, Fireworks had already crossed $1 billion in annualized revenue run rate, a fivefold year-over-year increase.

What

Atreides Management, Index Ventures, and TCV led the round. Nvidia, Lightspeed Venture Partners, Evantic Capital, Bessemer Venture Partners, Menlo Ventures, Insight Partners, Ontario Teachers' Pension Plan, and Lone Pine Capital also joined, per Quartz.

Token volume on the platform reached more than 40 trillion per day, up from 15 trillion over the same period, per Quartz. More than 95 percent of those tokens came from models that customers had specialized on their own proprietary data, per the Fireworks blog.

CEO Lin Qiao co-founded the company in 2022 alongside six other former Meta engineers. Fireworks runs inference infrastructure for open-weight models including Llama, Mistral, DeepSeek, and Kimi K3. Named customers include Cursor (coding models), Harvey (legal AI), Uber, Shopify, Doximity, Elastic, GitLab, and MongoDB, per Quartz. A $250 million raise at a $4 billion valuation preceded this round, per Quartz.

Why it matters

Qiao put the cost advantage in concrete terms: "Our cost compared with the equivalent-quality closed model is five to 10 times cheaper," she told CNBC, per Quartz. That gap is the specific reason customers like Cursor and Harvey built on Fireworks rather than routing traffic through closed-model APIs. At high daily token volumes, a 5-to-10x difference in per-token cost compounds quickly.

Nvidia's participation matters beyond its size. By joining the round, the chip maker now holds a direct stake in the success of the inference layer running at scale on its hardware, separate from relationships with any single model lab. Shared economic interest tends to speed up access to optimized kernels and early model support.

Valuation trajectory tells its own story: Fireworks jumped from a $4 billion valuation to $17.5 billion in under a year, per Quartz. Together AI raised $800 million in the same month. Two major inference infrastructure providers raising at this pace in a single month confirms institutional investors are treating specialized inference as a category distinct from model development, not a transitional layer that hyperscalers will absorb.

One caveat: the 95 percent specialized-token figure and the ARR are self-reported by Fireworks and have not been audited by a third party.

What to watch next

Qiao told Quartz she plans to triple the company's workforce of roughly 200 before year end. A hiring ramp that steep will test whether deployment quality and support keep pace with new customer onboarding. On the product side, whether Fireworks adds recently released open-weight models such as Kimi K3 to its supported catalog in the next 30 days, and how its per-token pricing evolves relative to Together AI following both firms' capital raises, are the two near-term markers that will show how the specialized inference market settles.

Sources