ARR Leaderboard
fal.ai went from roughly $95M annualized revenue in mid-2025 to ~$400M by early 2026, on usage-based billing for hosting 600+ generative media models. Every reported ARR, funding and valuation number is collected below, each traced back to a public source.
~$400M
Estimated ARR, early 2026
~$33.3M
Implied monthly recurring revenue
1,000%+
Implied annual growth
$4.5B
Valuation, Series C
Estimated annualized run rate in USD millions, mid-2025 to early 2026. Linear scale — the curve is nearly vertical, which is the point.
~$35M → ~$95M ARR
~6 months
2024 → Jul 2025
~$95M → ~$300M ARR
~3 months
Jul 2025 → Oct 2025
~$285M → ~$400M ARR
~2 months
Dec 2025 → Feb 2026
Implied annual growth
1,000%+
ARR Club: +1,259% YoY
2024
$35M ARR
Around $35M annualized revenue as the go-to serving layer for open-source image models like FLUX. — ARR Club
July 2025
$95M ARR
Annualized revenue around $95M as video model serving starts compounding on top of image workloads. — Sacra
October 2025
$300M ARR
Annualized revenue roughly tripled from July to ~$300M; Sacra puts end-of-2025 at ~$285M, up 1,040% YoY. — Sacra / ARR Club
Late 2025
$225M ARR
$125M Series C at a $4.5B valuation — tripled from the prior round — with ARR doubling again since February. — Notable Capital
February 2026
$400M ARR
Annualized revenue doubled to ~$400M since October, with fal reportedly targeting a raise near an $8B valuation. — Dealroom
Latest estimated ARR
~$400M
Dealroom, Feb 2026
Implied MRR
~$33.3M / month
Derived from ARR
Valuation
$4.5B, targeting ~$8B
Series C / Dealroom
Series C raised
$125M
Notable Capital
Models hosted
600+
fal.ai platform
Revenue basis
Usage-based inference billing
Sacra
Founded
2021
Company
fal.ai sells inference, not seats. Developers pay per request — per image generated, per second of video rendered, per audio output — through metered API billing with volume discounts, on top of dedicated capacity plans for high-volume customers. fal hosts 600+ third-party and in-house optimized models (FLUX, Wan, and others), so revenue scales directly with tokens of media generated rather than with headcount of customers.
The economics are a spread business: fal's low-latency serving infrastructure reportedly runs media model workloads up to 10x more efficiently than stock deployments, and that efficiency gap is the margin. Because billing is pure consumption, revenue tracks the virality of customer apps — which is why the curve can double in two months, and why the company is a rare inference platform that grew into hundreds of millions of ARR without an enterprise sales team.
Revenue Mix
FAQ
Per-request revenue needs per-request payments. Clink meters image, video and audio inference billing with merchant-of-record tax handling built in.
Meter Every Request→