Fireworks just hit $17.5B valuation. 95% of its tokens are now custom models.

Fireworks raised $1.5B Series D at $17.5B valuation.

9 months ago it was worth $4B.

The real number isn't the valuation. It's the revenue: $1B ARR, up 5x year-over-year.

And here's the stat that should keep every CIO awake tonight.

95% of the 40 trillion tokens Fireworks serves daily come from models trained on customer proprietary data.

Not general-purpose GPT wrappers. Custom models built on proprietary data for specific business functions.

Uber. Shopify. Samsung. Cursor. Harvey. MongoDB.

They're not renting intelligence anymore. They're owning it.

Lin Qiao, CEO: "There are two paths forward for AI. In one, intelligence belongs to a few big labs, and everyone else rents it. In the other, every company builds specialized intelligence of its own."

The $1B ARR proves the second path is winning.

Open models have reached parity with closed frontier models on specialized tasks. The cost advantage is 3-5x. The data advantage is infinite.

If your AI strategy still centers on OpenAI API calls, you're competing against companies that own their intelligence stack.

Audit your AI architecture this quarter. Ask one question: do we own our models, or do we rent them?

The answer determines who wins the next 3 years.

SOURCE: https://fireworks.ai/blog/series-d-announcement
VERIFIED: Fireworks.ai blog, Reuters, CNBC, BusinessWire
SIGNAL: The enterprise AI market just shifted from renting to owning. Companies building custom models on proprietary data are pulling away from those still calling the OpenAI API.

---