ARR Leaderboard
Modal turned per-second serverless billing for GPUs into roughly $300M of annualized revenue in about four years. Every reported ARR, funding and valuation number below is traced back to a public source — with the caveat that Modal itself has never published a revenue figure, so the latest number is a research estimate.
~$300M
Estimated ARR, April 2026
~$25M
Implied monthly recurring revenue
$4.65B
Valuation, May 2026 round
~6x
Annualized revenue growth in 12 months
Estimated and reported annualized run rate in USD millions, 2024 to 2026. Modal does not disclose revenue; figures are press reports and third-party estimates.
$0 → $50M ARR
~3 years
2022 founding → mid-2025
$50M → $119M ARR
~6 months
Mid-2025 → Dec 2025
$60M → $300M ARR
~7 months
Sep 2025 → Apr 2026
$1.1B → $4.65B valuation
~8 months
Sep 2025 → May 2026
2024
$10M ARR
Early revenue as the serverless GPU platform finds product-market fit with open-model inference workloads. Third-party estimate. — Sacra (estimate)
Mid-2025
$50M ARR
Around $50M annualized revenue as Modal is reported raising at a multi-billion valuation. — Business Insider
September 2025
$60M ARR
~$60M annualized revenue alongside an $87M Series B led by Lux Capital at a $1.1B post-money valuation. — Reuters / Lux Capital
December 2025
$119M ARR
~$119M annualized by year-end — roughly doubling in six months on inference and AI-agent sandbox demand. — Sacra
April 2026
$300M ARR
Sacra estimates ~$300M annualized revenue as AI coding-agent workloads explode on the platform. — Sacra
May 2026
$300M ARR
$355M new round at a $4.65B valuation, driven partly by AI coding workloads. Total funding ~$466M. — Reuters
Latest estimated ARR
~$300M
Sacra estimate, Apr 2026
Implied MRR
~$25M / month
Derived from ARR
Valuation
$4.65B
Reuters, May 2026
Total raised
~$466M
Tracxn
Founder & CEO
Erik Bernhardsson
Company
Modal sells fully metered serverless compute: functions run on demand in containers and customers pay per second of GPU or CPU time — from small A10 and L40S instances up to H100s — with no clusters to manage, no minimum commitments and idle capacity billed at nothing. Published pricing is per-second on each GPU class, which makes Modal the default way to serve open-weight model inference without negotiating a cloud contract.
The second engine is agentic workloads: Modal's sandbox primitives let AI coding agents spin up isolated environments thousands of times per minute, and batch, fine-tuning and training jobs ride the same per-second metering. Because revenue tracks actual compute consumed rather than seat counts, Modal's top line scales directly with its customers' token and agent volume — the pattern behind the ~6x jump from mid-2025 to early 2026.
Revenue Mix
FAQ
Per-second GPU billing is unforgiving — your metering has to match. Clink pairs usage-based subscriptions with checkout and merchant-of-record tax for serverless platforms.
Meter Your Compute→