$1B/year AWS Lambda for GPUs
Jan-Erik Asplund
TL;DR: Modal automatically scales GPU capacity with demand & charges by the second, letting AI companies run models without managing servers or paying for idle capacity between traffic spikes. With the rise of sandboxes used to run, train & evaluate AI agents, Sacra estimates Modal hit $1B in annualized revenue in October 2026. For more, check out our full report and dataset on Modal.
Key points via Sacra AI:
- Where AWS Lambda (2014) popularized serverless as a way to run code without managing web servers, Modal (2021) launched to do the same for deploying AI models on GPUs, with growth taking off as early generative AI companies like Mage Space (first web UI for Stable Diffusion) hit 1M users in 6 months (with Mage Space alone driving 80% of Modal’s 2023 revenue). Modal saves its customers (Suno, Runway, Decagon) money on spiky inference needs by buying compute from neoclouds and chopping & selling it by the second so they can spin GPUs up and down as needed instead of renting dedicated, always-on capacity.
- Launched in 2023 to give companies secure cloud environments where they could run LLM-generated code, Modal Sandboxes grew from near-zero revenue (May 2025) to $300M annualized (October 2026) as sandboxes became core to AI agent harnesses (Ramp, Linear), post-training (Cognition, Meta) & evals (Legora, Snorkel). Reusing the container runtimes & infrastructure they built for inference, Modal, Baseten ($600M annualized revenue, March 2026), RunPod ($325M annualized revenue in October 2026) & Fireworks ($1B annualized revenue in June 2026) are all expanding into sandboxes, going head-to-head with each other & with sandbox companies like E2B and Daytona, which are themselves adding GPUs for inference, training & reinforcement learning.
- Capitalizing on both the agent boom driven by smarter frontier models as well as the anti-tokenmaxxing shift towards cost-effective open-weights inference, Sacra estimates Modal hit $1B in annualized revenue in October 2026, up from $160M at the end of 2025, valued at $15.75B as of its October Series D (Accel) for a 15.8x revenue multiple. Compare to Fireworks AI at $1B ARR in June 2026, up 5x YoY, valued at $17.5B for a 17.5x multiple, Together AI at $1B in annualized revenue in February 2026, last valued at $8.3B as of its July Series C, and RunPod at $325M ARR in October 2026, up from $120M in January, last valued at $1B as of its June round for a 4.2x multiple on its then-$240M ARR.
For more, check out this other research from our platform:
- Modal Labs (dataset)
- RunPod (dataset)
- Baseten (dataset)
- Fireworks AI (dataset)
- Fal.ai (dataset)
- Replicate
- Together AI (dataset)
- Fluidstack
- Crusoe
- E2B
- Daytona
- RunPod customer at Segmind on GPU serverless platforms for AI model deployment
- No side quests for Lambda
- Fal.ai at $95M/year growing 4,650% YoY
- Together AI: the $44M/year Vercel of generative AI
- Fireworks AI customer at Hebbia on serving state-of-the-art models with unified APIs
