Fireworks secures $1.5B to build specialized AI
▼ Summary
– Fireworks raised $1.5 billion in Series D funding, valuing the company at $17.5 billion, to support its thesis that companies will build their own specialized AI models rather than rent them from large labs.
– The company runs and tunes open-source models on client data to create “specialized intelligence,” and its annualized revenue has surpassed $1 billion, up fivefold in a year.
– Fireworks processes over 40 trillion tokens daily, handling more requests than Google or OpenAI serve to developers, though it is smaller in revenue.
– The shift toward cheaper, capable open models is driving demand, with Fireworks claiming its option costs a fifth to a tenth of alternatives.
– Founded by former Meta engineers in 2022, Fireworks plans to grow headcount from 200 to 600 by year-end, facing competition from Together AI and Baseten.
The AI industry has long operated on a straightforward formula: rent a model from a major lab. Fireworks just raised $1.5 billion to champion the opposite vision,that every company will eventually build its own specialized intelligence.
The Series D round values the San Mateo startup at $17.5 billion, the company announced. Atreides Management, Index Ventures, and TCV led the investment, with existing backer Nvidia also participating.
“Fireworks has assembled one of the most elite and technical teams in AI,” said Gavin Baker, Atreides’ managing partner, in the funding announcement.
What Fireworks sells is straightforward but powerful. It runs open models for other organizations, then helps them fine-tune those models on proprietary data. The result is what the company calls “specialized intelligence”,a model infused with the unique knowledge only one business possesses.
The thesis is gaining traction. Revenue has surpassed $1 billion on an annualized basis, a fivefold increase in a single year, according to Index Ventures. The platform now processes more than 40 trillion tokens daily, with 95% of that traffic running on specialized models.
In one key metric, Fireworks is already bigger than the tech giants. The company handles more requests each day than either Google or OpenAI report serving to developers, based on CNBC’s numbers. While its revenue is a fraction of theirs, its traffic volume is larger.
That demand reflects a broader industry shift. As open models approach parity with the best closed ones, the AI race has moved from building the biggest model to finding the cheapest capable one. Fireworks claims its option runs at one-fifth to one-tenth the cost.
The argument has powerful allies. CEO Lin Qiao framed the choice plainly when announcing the round. “In one path, intelligence belongs to a few big labs, and everyone else rents it,” she said. “We are building towards the second.”
She is not alone. Microsoft’s Satya Nadella argues that a firm should use a model without surrendering the knowledge that makes it unique. Palantir’s Alex Karp says clients want to “own the means of production.”
The market is crowded and getting cheaper. Former Meta engineers behind PyTorch founded Fireworks in 2022. It competes with Together AI and Baseten on inference, and increasingly with neocloud rivals on training.
Its biggest vulnerability is concentration. Cursor once accounted for roughly half its revenue, though Qiao says the customer base is now more diversified. She plans to grow headcount from 200 to 600 by year-end. “This is the year when we’ll really hit the gas,” she told CNBC.
(Source: The Next Web)
