Fireworks has secured $1.5 billion to develop specialized intelligence.
The AI economy has followed a straightforward pattern: companies rent models from large laboratories. Fireworks has recently raised $1.5 billion to challenge this notion, arguing that every organization will develop its own intelligence. The Series D funding values the San Mateo startup at $17.5 billion, as stated by the company. The funding round was led by Atreides Management, Index Ventures, and TCV, with Nvidia, a prior investor, participating once again.
“Fireworks has brought together one of the most talented and technical teams in AI,” remarked Gavin Baker, managing partner at Atreides, during the funding announcement.
What Fireworks provides
Fireworks operates open models for various companies and assists them in refining these models using their specific data. This process results in what it terms "specialized intelligence," meaning a model tailored by the unique insights of a single business.
The latest buzz in EU tech, insights from our founder Boris, and some intriguing AI art can be found in our weekly newsletter. It's free, so sign up now! The company’s concept is gaining traction. According to Index Ventures, revenue has exceeded $1 billion on an annual basis, increasing fivefold within a year. The platform now handles over 40 trillion tokens daily, with 95% utilizing specialized models.
Larger than the labs, in some respects
That token statistic is impressive. Based on CNBC’s data, Fireworks processes more daily requests than Google or OpenAI report serving to developers. While its revenue is only a fraction of theirs, it generates more traffic.
This demand reflects a broader trend. As open models improve and close the gap with leading closed models, the AI competition is shifting focus from the largest model to the most cost-effective one. Fireworks claims that its option operates at only 20% to 10% of the cost.
The argument has influential backers
CEO Lin Qiao presented it as a clear choice during the funding announcement. “In one scenario, intelligence resides with a few large labs, and everyone else rents it,” she explained. “We are advancing towards the alternative.”
She is not alone in this belief. Microsoft's Satya Nadella argues that firms should leverage a model while retaining the unique knowledge that sets them apart. Palantir’s Alex Karp emphasizes that clients prefer to “own the means of production.”
A competitive, lower-cost market
Fireworks was founded in 2022 by former Meta engineers behind PyTorch. It competes with Together AI and Baseten in inference while increasingly challenging neocloud competitors in training.
Its main risk is dependency on a few sources of revenue. Cursor previously accounted for about half of its income, although Qiao asserts that the revenue base is now more diverse. She aims to increase the workforce from 200 to 600 by the end of the year. “This is the year we intend to accelerate significantly,” she stated to CNBC.
Other articles
Fireworks has secured $1.5 billion to develop specialized intelligence.
Fireworks secured $1.5 billion at a valuation of $17.5 billion, believing that every company will develop its own tailored intelligence rather than leasing it from a major laboratory.
