Fireworks AI raises $1.5B at $17.5B valuation for open model infrastructure

Fireworks AI has raised $1.505 billion in Series D funding at a $17.5 billion valuation, a new financing for the enterprise AI infrastructure company founded by former Meta engineering leader Lin Qiao. Atreides Management, Index Ventures and TCV led the round, with participation from NVIDIA, Lightspeed Venture Partners, Evantic, 20VC, Bessemer Venture Partners, Insight Partners, Lone Pine Capital, Menlo Ventures, Ontario Teachers’ Pension Plan and others.

The company said it has passed $1 billion in annualized revenue, up fivefold from a year earlier, while daily token volume on its platform has risen from 15 trillion to more than 40 trillion. Fireworks did not specify whether the reported valuation is pre- or post-money.

Fireworks sells infrastructure for companies that want to fine-tune and run open-source AI models using their own data, instead of relying only on general-purpose models from large AI labs. The company says 95% of tokens processed on its platform now come from customer-customized models rather than standard frontier models. Named customers include Uber, Shopify, Revolut and Doximity.

Qiao’s background is central to Fireworks’ pitch. Before founding the company in 2022, she spent seven years at Meta, from July 2015 to September 2022, and led more than 300 engineers working on PyTorch, the open-source machine learning framework used across much of the AI industry, including at OpenAI, Google and Meta. Fireworks was co-founded with five other former Meta engineers, including Dmytro Dzhulgakov and Dmytro Ivchenko, who were PyTorch core maintainers at Meta before joining the startup.

Revenue growth comes with concentration questions

The financing follows a period in which Fireworks had to address customer concentration. As of last year, about half of its revenue came from Cursor, the AI coding company. That exposure became more relevant after SpaceX announced a $60 billion all-stock deal to acquire Cursor, a transaction that had not closed and was expected to require regulatory approval in the third quarter of 2026.

Cursor has started scaling its Composer model on SpaceXAI’s Colossus compute infrastructure, reducing its reliance on outside inference providers such as Fireworks. Qiao has said Fireworks is now more diversified, and the company points to customers including Uber, Shopify and Revolut. The company has not disclosed how much of the fivefold revenue increase came from Cursor or from newer enterprise accounts.

Inference infrastructure is getting crowded

Fireworks is raising into one of the best-funded parts of AI infrastructure. Together AI recently raised $800 million at an $8.3 billion valuation. Baseten raised $300 million at a $5 billion valuation from IVP and CapitalG, with NVIDIA contributing $150 million. Groq raised $650 million in June 2026 after licensing its chip technology to NVIDIA.

The bigger competitive pressure may come from the cloud platforms. Microsoft Azure and AWS are adding managed open-model hosting to their own services, which could put similar capabilities inside existing enterprise cloud contracts. NVIDIA is also both an investor in Fireworks and an increasingly active player in inference through its own expansion.

Fireworks is positioning its differentiation around model customization, arguing that enterprises need models adapted to domain-specific data rather than faster hosting alone. That claim will be tested as hyperscalers bundle more AI tooling into their platforms.

The company said it will use the new capital for global compute expansion, engineering hiring and deeper partnerships with Microsoft and NVIDIA. Fireworks expects headcount to rise from about 200 employees to 600 by the end of 2026.

The Series D follows earlier financings that valued Fireworks at $552 million in its Series B in July 2024 and $4 billion in its Series C in October 2025. Grand View Research has estimated the generative AI market at $22.2 billion in 2025 and projected it could reach $324.7 billion by 2033, but Fireworks’ nearer-term test is whether open-model customization remains a standalone market as cloud incumbents move further into the same category.

This story draws on original reporting from TechFundingNews.