ARR Leaderboard
Baseten roughly tripled its annualized revenue from ~$200M to ~$600M inside a single quarter at the turn of 2026, then closed a $1.5B round at a $13B valuation. Every reported revenue, funding and valuation number below is traced back to a public source.
~$600M
Annualized revenue, March 2026
~$50M
Implied monthly recurring revenue
$13B
Valuation, Series F Aug 2026
~3x
ARR growth in a single quarter
Reported annualized run rate in USD millions, December 2025 to August 2026. Linear scale — which is why the pre-2026 curve barely registers.
$0 → $200M ARR
~7 years
2019 founding → Dec 2025
$200M → $600M ARR
~3 months
Dec 2025 → Mar 2026
$11B → $13B valuation
Months
Early 2026 → Aug 2026
Capital raised in 12 months pre-Series F
~$525M
Three rounds, 2025–2026
2024
$25M ARR
Early traction on model-serving infrastructure after the Series C era. Third-party estimate — Baseten does not publish revenue. — Sacra (estimate)
December 2025
$200M ARR
Annualized revenue of roughly $200M as frontier-model serving demand surges. — Value Add VC / Sacra
March 2026
$600M ARR
~$600M annualized revenue — roughly tripling in a single quarter, an estimated ~19x year over year. — The Information / Value Add VC
Early 2026
$600M ARR
The Information reports Baseten in talks to raise ~$1B at an $11B valuation as Q1 revenue surges. — The Information
August 2026
$600M ARR
$1.5B Series F closed at a $13B valuation — one of the largest AI inference rounds ever — led by Altimeter and others. — Company / press reports
Latest reported revenue
~$600M annualized
The Information, Mar 2026
Implied MRR
~$50M / month
Derived from ARR
Valuation
$13B
Series F, Aug 2026
Series F
$1.5B
Altimeter and others, Aug 2026
Growth rate
~19x YoY
Value Add VC / Sacra
Baseten sells dedicated model-serving infrastructure priced by GPU-minute: customers reserve and consume GPU capacity (throughput-optimized for production serving) to run frontier and open-weight models behind predictable, production-grade SLAs. The load sits with enterprises and AI startups that need inference performance beyond what per-token APIs from labs offer — custom weights, custom latency, custom autoscaling.
The revenue base is concentrated among large AI consumers, which is exactly what made a 3x quarter possible: when a handful of major customers scale serving volume, committed GPU-minute spend scales with it. Analysts flag the flip side — GPU scarcity and the cost of dedicated capacity pressure margins, which is why the $13B valuation drew scrutiny alongside the $600M figure.
Revenue Mix
FAQ
GPU-minute pricing needs metering, not invoices. Clink adds usage-based billing, committed contracts and merchant-of-record tax handling for compute-heavy serving platforms.
Meter GPU Minutes→