Fireworks raises $1.5B at $17.5B valuation as AI infrastructure demand surges

Nvidia-backed Fireworks surpassed $1B in annualized revenue and raised $1.5B as enterprises shift from frontier models to customized open-source AI.

Fireworks, the AI infrastructure startup backed by Nvidia, surpassed $1 billion in annualized revenue — up fivefold from a year earlier — as companies increasingly customize open-source models rather than renting frontier AI from big labs.

"There are two paths forward for AI. In one, intelligence belongs to a few big labs, and everyone else rents it. In the other, every company in the world builds specialized intelligence of its own," Lin Qiao, co-founder and chief executive officer of Fireworks, said.

The company now serves more than 40 trillion tokens daily, nearly triple the 15 trillion it handled during its last funding round. Ninety-five percent of tokens processed on Fireworks come from models that have been customized, the company said. The Series D round, led by Atreides Management, Index Ventures and TCV, values Fireworks at $17.5 billion — more than four times its $4 billion valuation from October.

The fundraising reflects a structural shift in enterprise AI strategy. Companies including Uber, Shopify and Doximity are using Fireworks to train open models on proprietary data, creating what Qiao calls "specialized intelligence" that can match or beat closed models on specific tasks at lower cost. Fireworks said its pricing is five to 10 times cheaper than equivalent closed models.

Fireworks competes in the inference cloud market alongside startups such as Together AI and Baseten, and has expanded into GPU provisioning for training, putting it in competition with neocloud providers CoreWeave, Lambda and Nebius. CoreWeave, which raised $1.5 billion in an initial public offering last year, is now valued at $42 billion.

The company's customer base has diversified significantly. As of last year, about half of Fireworks' revenue came from AI coding startup Cursor, which SpaceX agreed to acquire in June for $60 billion in stock. "We are much more diversified right now," Qiao said. Other clients include Elastic, GitLab and MongoDB.

In March, Fireworks announced a partnership with Microsoft, allowing customers of the Windows and Office company to draw on models through Fireworks' platform, which relies on computing power from more than 20 suppliers. "Through Microsoft we can get much bigger reach," Qiao said.

Fireworks employs about 200 people and expects to reach 600 by the end of 2026. The company hired former Salesforce executive George Hu as its president in April and plans to build a dedicated sales team after years of relying on self-serve customer sign-ups. The new capital will also fund additional GPU capacity and engineering hires.

The funding round included participation from Nvidia, Lightspeed Venture Partners, Bessemer Venture Partners, Insight Partners, Menlo Ventures and Ontario Teachers' Pension Plan, among others. Nvidia, which trades at about 35 times forward earnings, has invested in multiple AI infrastructure startups as it seeks to broaden the ecosystem for its GPUs beyond the handful of large cloud providers. Fireworks' daily token volume of 40 trillion now rivals Google's disclosed rate of about 27 trillion tokens per day and OpenAI's roughly 22 trillion.

This article is for informational purposes only and does not constitute investment advice.