The AI infrastructure provider Fireworks AI has completed a Series D funding round of $1.5 billion, resulting in a valuation of $17.5 billion. The valuation has quadrupled within eight months compared to the previous round. The funding was led by investors Atreides Management, Index Ventures, and TCV.
Series D Round Quadruples Valuation in Eight Months
Fireworks AI raised $250 million in a Series C round in November 2025 at a valuation of $4 billion. The current round now brings the company an additional $1.5 billion in capital. In addition to the leaders Atreides Management, Index Ventures, and TCV, existing investors such as Nvidia, Lightspeed Venture Partners, and Evantic Capital, as well as more than half a dozen other new backers, participated according to company statements. The company plans to invest the capital in expanding its computing infrastructure and in new engineering positions. Fireworks AI was founded in 2022 by a team of former Meta and Google employees. CEO Lin Qiao previously led the PyTorch development at Meta, and her co-founders worked on Meta’s advertising infrastructure and Google’s Vertex AI platform, among other projects. With the Series C round, total funding amounted to approximately $327 million; after the current round, the total capital raised exceeds $1.8 billion. Nvidia’s participation underscores the close interconnection between chip and infrastructure providers in the current AI market: The chip manufacturer has invested in several inference specialists in recent months to secure its chip demand in the long term.
Revenue and Token Volume Grow at Record Pace
According to company statements, Fireworks AI’s annual revenue rate exceeds $1 billion for the first time, a fivefold increase year-over-year. The platform now processes more than 40 trillion tokens per day, compared to around 15 trillion at the beginning of the year. More than 95 percent of these tokens, according to Fireworks, come from models that customers have tailored to their own data, rather than from unchanged standard models. These revenue figures are independently unverified. Among the mentioned customers are the programming platform Cursor and the legal service provider Harvey. The platform offers more than 200 open models for text, image, and other formats and claims to support new open-source releases shortly after their appearance. The strong increase in token volume reflects a broader trend: More and more companies are now using AI applications productively rather than just experimentally, which increases the demand for fast and cost-effective inference capacity. The platform also includes an agent that automatically optimizes hyperparameters and expands training datasets with so-called DPO files – according to company statements, a way to improve model quality without additional manual effort.
Fireworks Positions Itself in the Competition of Inference Providers
Fireworks AI competes with providers like Together AI and Baseten for companies that want to use their own AI models instead of ready-made chatbot products. Co-founder and CEO Lin Qiao justifies the strategy by stating that companies could develop an independent knowledge advantage from their own data and workflows that general models could not represent. The platform offers serverless and dedicated inference services, several methods for distributed training, and a system that automatically expands training data and compresses models. Other customers include, according to a report by Pulse2, Uber, Shopify, Samsung Electronics, and GitLab. Customers can also test multiple open models in parallel on the platform and have them automatically selected for individual tasks. The jump from $4 billion to $17.5 billion within a few months shows how strongly investors are currently betting on specialized AI infrastructure, while the valuations of large foundation model providers are increasingly being discussed critically. Unlike providers that develop their own base models, Fireworks consciously refrains from having its own foundation model and instead positions itself as an infrastructure layer for open models from different manufacturers. Fireworks did not provide a specific timeline for a potential IPO.
It remains to be seen whether Fireworks can maintain its revenue growth as cloud providers like AWS and Google Cloud continue to expand their own inference services. The company has not yet announced a date for an IPO or further large funding rounds.
