ARR Leaderboard
AI21 Labs — the Tel Aviv lab founded in 2019, older than most of this leaderboard — reached a ~$350M estimated annualized run rate in 2026 by selling efficient enterprise LLMs: the Jamba model family, the Maestro orchestration platform and the Wordtune writing product. Every reported figure is collected below, each traced back to a public source; AI21 has not officially disclosed revenue.
~$350M
Estimated ARR, 2026
~$29.2M
Implied monthly recurring revenue
$300M
Series D, May 2025
$1.5B
Valuation
Estimated annualized run rate in USD millions, 2023 to 2026. AI21 has never officially disclosed revenue — all points are third-party estimates.
$0 → $10M ARR
~4 years
2019 founding → ~2023
$20M → $50M ARR
~18 months
2023 → mid-2025 (est.)
$50M → $350M ARR
~12 months
2025 → 2026 (est.)
Revenue multiple 2023→2026
~17x
$20M → ~$350M (est.)
2019
$0M ARR
Founded in Tel Aviv by Yoav Shoham, Ori Goshen and Amnon Shashua — one of the earliest dedicated LLM companies, predating the current wave by four years. — Company
November 2023
$20M ARR
~$208M Series C at a ~$1.4B valuation, with revenue estimated in the low tens of millions from the Wordtune consumer product and early enterprise API deals. — AI21 / press reports
2024
$30M ARR
The Jamba family — the first production hybrid SSM-Transformer LLMs — ships in 2024, opening long-context enterprise workloads; revenue estimated in the tens of millions. — Company / press reports
May 2025
$50M ARR
$300M Series D from Google and Nvidia; third-party estimates put 2025 revenue around $50M as the company pivots toward Maestro, its enterprise AI-orchestration platform. — Forge Global / PM Insights
2026
$350M ARR
Third-party estimates put the annualized run rate near $350M as Jamba-based enterprise deployments and Maestro scale — a figure AI21 itself has not confirmed. — Presenc AI
Latest estimated ARR
~$350M
Presenc AI, 2026
Implied MRR
~$29.2M / month
Derived from ARR
Valuation
$1.5B
Series D, May 2025
Total raised
~$636M
Across 7 rounds, May 2025
Founded
2019, Tel Aviv
Company
AI21's model is enterprise efficiency: the Jamba family (hybrid SSM-Transformer architecture) is sold on a usage-based API with per-token pricing positioned well below frontier-lab rates, optimized for long-context, high-throughput workloads where cost per million tokens is the buying criterion. Jamba models are also offered open-weight for private deployment, mirroring the sovereign/enterprise playbook.
Two adjacent products round out the mix: Maestro, an agentic orchestration platform that composes AI21 and third-party models into enterprise workflows (the strategic focus of the 2025 Google/Nvidia-backed round), and Wordtune, the consumer writing assistant that was AI21's first product and still contributes a consumer revenue base.
Revenue Mix
FAQ
Clink is the payment rails for token-metered enterprise APIs — usage billing, committed spend and merchant-of-record tax handling built for cost-per-million-token businesses.
Price Your Tokens→