Fireworks AI, a cloud platform that gives developers access to managed GPU clusters for customizing and running open-source AI models, has closed a $1.5 billion Series D round at a $17.5 billion valuation. Atreides Management, Index Ventures and TCV led the round, with Nvidia also participating.
The company's annualized revenue recently crossed $1 billion, and its platform processes more than 40 trillion tokens per day. Customers include Samsung Electronics and GitLab.
Fireworks AI's platform lets engineering teams customize open-source models without managing the underlying infrastructure. An automated agent handles training workflows, including hyperparameter optimization and dataset preparation, while developers define task objectives and supply training data. The platform supports four parallelization techniques that can run concurrently or individually depending on model architecture, and offers both serverless inference environments and a dedicated Deployments service with autoscaling and model compression through quantization.
Lin Qiao, co-founder and CEO of Fireworks AI, described the commercial thesis behind the platform: "Every company possesses unique proprietary knowledge — their data, operational processes, customer insights, and quality standards. Our mission at Fireworks is to transform this distinct knowledge into specialized, proprietary intelligence that organizations can continuously refine and leverage."
The company said it will use the capital to expand infrastructure capacity and grow its engineering team.
