Inference-layer revenue run-rates
Annualised revenue, bookings and run-rates for inference clouds, serving platforms and routers (Fireworks, Baseten, Together, Modal, RunPod, OpenRouter, Groq), hand-entered from company statements, press reports and Sacra's estimates, each checked against its page.
Reason through thisThe reading
Several inference platforms report run-rates between a few hundred million and about a billion dollars, growing several-fold in a year, while the price of a token at fixed capability falls; revenue says nothing about their margins, which none publishes. The direction reads Fireworks through Sacra's estimates, one estimator for one firm, so the other firms stay as evidence rather than votes.
2 readings drawn from sacra_free.fireworks.revenue_run_rate_usd.pt, one of 15 series listed under How it is measured; 21 observations in the evidence.
Source: Sacra company pages (free summaries). Sacra estimates (free company page).
Why it matters
The serving layer sits between the model makers and the applications; whether it keeps margin as token prices fall decides whether it is a durable business or a pass-through.
What cuts against it
No figure is audited; bookings, run-rates and net router fees are different measures; OpenRouter's figure is its fee on spend, not the spend; Groq sits in the chips layer.
How it is measured
- series
sacra_free.fireworks.revenue_run_rate_usd.pt - series
fireworks_blog.fireworks.revenue_run_rate_usd.pt - series
cnbc.fireworks.revenue_run_rate_usd.pt - series
pymnts.baseten.revenue_run_rate_usd.pt - series
sacra_free.baseten.revenue_run_rate_usd.pt - series
techcrunch.together.annual_bookings_usd.pt - series
sacra_free.together.revenue_run_rate_usd.pt - series
crunchbase_news.together.revenue_run_rate_usd.pt - series
yahoo_finance.modal.revenue_run_rate_usd.pt - series
techcrunch.modal.revenue_run_rate_usd.pt - series
runpod_press.runpod.revenue_run_rate_usd.pt - series
thenextweb.runpod.revenue_run_rate_usd.pt - series
pymnts.openrouter.revenue_run_rate_usd.pt - series
sacra_free.openrouter.revenue_run_rate_usd.pt - series
sacra_free.groq.revenue_usd.fy - source CNBC · default tier 5 · CNBC; short quotation
- source Crunchbase News · default tier 5 · Crunchbase; short quotation
- source Fireworks AI blog · default tier 5 · Fireworks AI; short quotation
- source PYMNTS · default tier 5 · PYMNTS; short quotation
- source Runpod press releases · default tier 5 · Runpod; short quotation
- source Sacra company pages (free summaries) · default tier 6 · Sacra; short quotation of the free summary
- source TechCrunch · default tier 7 · TechCrunch; short quotation
- source The Next Web · default tier 5 · The Next Web; short quotation
- source Yahoo Finance (syndicated Investing.com) · default tier 5 · Syndicated; short quotation
- proxy capital_flow, market_structure · cadence irregular
The rule compares three estimates from one estimator (Sacra) for the largest serving firm; moves under $100M are estimation noise. Other publishers reporting the same firm (the Fireworks blog, CNBC) are the second source. Rising run-rate means the serving layer is keeping more of what passes through it.
Applied to sacra_free.fireworks.revenue_run_rate_usd.pt.
Latest readings
- 2026-05-31 fireworks · $800M⚑disputed: An annualised month reported by the press or the company, not booked or audited revenue, and often sourced to the company itself.run-rateas of 2026-05-31
- 2025-12-31 fireworks · $305M⚑disputed: An annualised month reported by the press or the company, not booked or audited revenue, and often sourced to the company itself.run-rateas of 2025-12-31
Why this status, and its history
First reading. Sacra estimates Fireworks, the largest serving firm the site tracks, at an annualised run-rate of $305M at the end of 2025 and $800M at the end of May 2026. Its direction rule compares three readings from the one estimator, so with two it reads emerging; the other firms (Together, Baseten, Modal, RunPod, OpenRouter) stay as evidence.
- 2026-09-27unmeasured to emergingconf — → 35 · Claude (model)
First reading. Sacra estimates Fireworks, the largest serving firm the site tracks, at an annualised run-rate of $305M at the end of 2025 and $800M at the end of May 2026. Its direction rule compares three readings from the one estimator, so with two it reads emerging; the other firms (Together, Baseten, Modal, RunPod, OpenRouter) stay as evidence.
Weighed against it: Run-rates are estimates annualised from one month; no serving firm publishes margins, so revenue growth says nothing yet about whether the layer keeps its rent.
Why this confidence
Rests on 21 observations from CNBC, Crunchbase News, Fireworks AI blog, PYMNTS, Runpod press releases, Sacra company pages (free summaries), TechCrunch, The Next Web, Yahoo Finance (syndicated Investing.com); best evidence tier 5, grade C; current.
Confidence is independent of status: 90–95 multiple strong independent sources; 70–89 good evidence, some ambiguity; 50–69 mixed or hard to operationalise; 0–49 limited or vague evidence.