READ. SCROLL. LISTEN.

Original briefings. Zero spin.

Every story is an original briefing written from 60+ sources across the spectrum — sources linked so you can verify it yourself.

← Back to headlines

Fireworks AI Hits $17.5 Billion Valuation as Companies Ditch Expensive Frontier Models

Fireworks AI Hits $17.5 Billion Valuation as Companies Ditch Expensive Frontier Models
Nvidia-backed Fireworks raised $1.51 billion at a $17.5 billion valuation on Thursday, cashing in on companies' desire to escape pricey subscriptions to OpenAI and Anthropic. Its revenue jumped fivefold in a year as businesses figure out they don't need to rent Cadillac AI when a tuned Honda does the job for a fraction of the cost. That's the free market working exactly how it's supposed to.

A cheaper AI bet just got a $17.5 billion price tag

Fireworks AI, a San Mateo startup that helps companies run open-source AI models instead of renting from the big labs, announced Thursday it raised $1.51 billion in a Series D round. That values the company at $17.5 billion, according to Reuters. Atreides Management, Index Ventures and TCV led the round, with Nvidia, Lightspeed Venture Partners, Bessemer Venture Partners, Insight Partners, Menlo Ventures, Ontario Teachers' Pension Plan and Lone Pine Capital also participating.

The company said it has crossed $1 billion in annualized revenue run rate, up fivefold from a year ago, according to Reuters. Daily token volume, the basic unit AI models chew through, jumped from 15 trillion to more than 40 trillion over that span, per Reuters and TNW's reporting on Index Ventures' figures.

Token volume is a real measure of usage, not hype. According to TNW, Fireworks now handles more daily token requests than Google or OpenAI report serving developers, based on figures cited by CNBC, even though its revenue is a fraction of theirs. Translation: it's not the biggest name in the room, but it's doing serious volume.

Why finance departments are tired of paying full price

CNBC reported that finance executives are increasingly anxious about the cost of frontier AI models from labs like OpenAI and Anthropic, both valued above $800 billion this year, and are pushing employees toward open-source alternatives instead.

Fireworks CEO Lin Qiao told CNBC the company is seeing "super-linear demand" and called it "a once-in-a-lifetime opportunity." She framed the strategic choice bluntly in the company's funding announcement, cited by TNW: "In one [path], intelligence belongs to a few big labs, and everyone else rents it. We are building towards the second."

Fireworks doesn't build its own foundation models from scratch. It hosts open-weight models, including releases from Chinese firms DeepSeek, MiniMax and Z.ai, plus open models OpenAI itself released last year, then helps businesses fine-tune those models on their own proprietary data. Qiao calls the output "specialized intelligence," as opposed to the "generalized intelligence" sold by the frontier labs, per CNBC.

That pitch runs at a fifth to a tenth of the cost of comparable closed models, according to TNW's reporting on the company's own claims. Customers include Uber, Shopify and Doximity, according to Reuters.

The bigger picture: cloud giants don't own this market

Fireworks' rise is also evidence that Amazon, Microsoft and Google haven't locked up cloud AI infrastructure the way some assumed they would. DigitalOcean stock is up 149% this year, CNBC noted, and CoreWeave, which rents out Nvidia GPUs, is now worth $42 billion after its 2025 IPO. Fireworks competes in this same inference-cloud lane alongside Together AI and Baseten, and is now pushing into GPU training infrastructure too, going head-to-head with neoclouds like CoreWeave, Lambda and Nebius.

Rather than fight Microsoft outright, Fireworks struck a partnership with the company in March, plugging into Microsoft's Foundry service so its customers can access Fireworks-hosted models. "Through Microsoft we can get much bigger reach," Qiao told CNBC.

A fair question: is this durable, or is Fireworks riding a temporary cost-arbitrage wave that closes once frontier labs cut their own prices? OpenAI and Anthropic have both slashed API pricing multiple times as competition intensified, and nothing stops them from doing it again. If the price gap narrows, the case for switching to open models weakens.

There's also a concentration risk TNW flagged. Cursor, the AI coding tool, once accounted for roughly half of Fireworks' revenue. Qiao says that customer base is now more diversified, but the company hasn't published a full breakdown, and a single-customer dependency at that scale is the kind of thing investors watch closely.

What happens next

Qiao told CNBC she plans to grow Fireworks' headcount from 200 to 600 employees by the end of 2026, calling this "the year when we'll really hit the gas." The company also said it will deepen partnerships with cloud players like Microsoft and Nvidia, according to Reuters.

Fireworks previously raised $250 million at a $4 billion valuation in October, per Reuters, meaning its valuation has more than quadrupled in under a year. Whether that holds depends on whether "specialized intelligence" stays cheaper than what OpenAI and Anthropic can offer once they respond to the pricing pressure.

Sources used for this briefing

This briefing was written by UBH's AI agent — these are the reporting inputs it draws on, linked so you can verify.

center-left
CNBCNvidia-backed Fireworks hits $17.5 billion valuation as companies pursue cheaper AI models
unknown
thenextwebFireworks raises $1.5bn to build specialized intelligence - TNW
unknown
whblNvidia-backed startup Fireworks valued at $17.5 billion in latest funding - WHBL