Fireworks AI is one of the fastest-scaling companies in AI infrastructure. Founded in 2022 by Lin Qiao — who previously led the PyTorch team at Meta — the Redwood City company runs an inference cloud that lets enterprises deploy, fine-tune, and scale open-source and custom AI models without managing GPU fleets themselves. It now serves more than 40 trillion tokens per day.
The numbers behind that growth are unusual even by 2026 AI standards. In July 2026, Fireworks announced a $1.505 billion Series D at a $17.5 billion post-money valuation, led by Atreides Management, Index Ventures, and TCV, with participation from Nvidia, Lightspeed, Bessemer, Menlo Ventures, Insight Partners, Ontario Teachers’ Pension Plan, and others. The round coincided with the company crossing $1 billion in annualized revenue — roughly five times its level a year earlier. Nine months before, its Series C had priced at $4 billion.
Can you buy Fireworks AI stock?
Not on any exchange. Fireworks AI is a private company — there is no ticker. The ways accredited investors get economic exposure are:
- Direct secondary purchases — buying outstanding shares from an existing shareholder, subject to company transfer restrictions and rights of first refusal.
- Special purpose vehicles (SPVs) — pooled vehicles holding Fireworks shares, letting investors participate at smaller check sizes than a direct block requires.
- Indirect exposure via investors — Nvidia and other public participants hold stakes, but the actual dollar exposure delivered through a public proxy is a small fraction of the headline ownership percentage.
For most accredited investors, the realistic path is the first two.
The Series D and where the valuation stands
The trajectory: a $52 million Series B in July 2024 at $552 million, a $250 million Series C in October 2025 at $4 billion, and the $1.505 billion Series D in July 2026 at $17.5 billion. A roughly 4x markup in nine months would normally signal a valuation running ahead of the business — here it tracked revenue almost exactly, with ARR growing from roughly $315 million in early 2026 to past $1 billion by mid-year.
What that means for secondary buyers: the last-round price is fresh and high. Secondary blocks in companies that just re-priced tend to trade near the new mark, sometimes at modest discounts from early employees seeking liquidity. The question isn’t whether you can find a discount to the Series D — it’s whether the growth persists long enough to make the Series D price cheap in hindsight.
Points of caution before extrapolating:
- Inference is a competitive, capital-intensive market — Baseten, Together, and the hyperscalers themselves all compete for the same workloads, and pricing pressure on inference is relentless.
- Usage-based revenue is durable but not contractually locked the way multi-year SaaS is; a customer can migrate workloads.
- At a $17.5 billion valuation, the round already prices in years of continued hypergrowth.
How access actually works
Fireworks AI secondaries trade through brokers and SPV syndicators, typically with minimums from $100,000 upward. Because the company is private, expect:
- Transfer restrictions. The company can block or delay secondary transfers; many trades settle through existing SPVs or as forward arrangements.
- Scarcity. Post-Series D, most holders have little urgency to sell; supply is thin and moves quickly when it appears.
- Information asymmetry. Work with intermediaries who disclose what is and isn’t confirmable.
At the time of writing, we have visibility into live Fireworks AI secondary supply, including allocations accepting indications from accredited investors. Terms, pricing, and structure are available on request to qualified investors.
Is Fireworks AI worth the risk?
The bull case: revenue quintupling year over year, a founder who built the infrastructure layer most of modern AI runs on, Nvidia on the cap table, and a customer base of production enterprise workloads rather than experiments. The bear case: a rich fresh mark, brutal competition in inference, dependence on continued open-model momentum, and the standard private-market risks — illiquidity, limited information, and no guaranteed exit.
This is a concentrated, illiquid position sized for the portion of a portfolio that can go to zero. That’s the honest description of every pre-IPO position, and Fireworks AI is no exception.