The AI compute platform — GPU cloud, 1-Click Clusters, and workstations for training and inference
HQ: San Francisco, CA, United States | Founded: 2012 | Employees: ~400 (2026, est.) | Stage: Late Stage VC (Series D+) | Website: https://lambda.ai
Lambda was founded in 2012 by brothers Stephen Balaban and Michael Balaban as a deep-learning hardware company selling GPU workstations and servers to researchers — long before the current AI boom. That early foothold gave Lambda deep relationships with the AI research community and the operational discipline to build Lambda Cloud, its on-demand GPU cloud launched in 2020, which became one of the most popular ways for startups and labs to rent H100-class compute.
Today Lambda operates three pillars: Lambda Cloud (per-second on-demand GPU instances including H100, H200, B200, and GB300 systems), 1-Click Clusters (self-serve production clusters from 16 to 2,000+ GPUs for training and fine-tuning), and the hardware business (GPU workstations and servers). It also runs Lambda Chat, an LLM chat product serving open models like Llama and DeepSeek with a 1M-token context window. The company positions itself as 'the world's least expensive GPU cloud,' competing on price and reliability against CoreWeave and the hyperscalers.
Financially, Lambda generated approximately $425 million in revenue in 2024 (Contrary Research) and has raised over $2.4 billion, including a $320M Series C at $1.5B (February 2024), a $480M Series D at $2.5B (February 2025, with NVIDIA among investors), and a $1.5B structured raise in November 2025 to fund what it calls the 'Superintelligence Cloud' — gigawatt-scale data centers. In November 2025 Lambda announced a multi-billion-dollar, multi-year agreement with Microsoft to deploy tens of thousands of NVIDIA GB300 GPUs, and in May 2026 it hired former Sprint and Alcatel-Lucent CEO Michel Combes as CEO, with co-founder Stephen Balaban moving to CTO to lead technology strategy.
Lambda is one of the strongest independent GPU cloud players, with a 12-year head start in deep-learning hardware (workstations and servers) that seeded one of the earliest and most reliable on-demand GPU clouds. It generated ~$425M in revenue in 2024, has raised over $2.4B, and has locked in a landmark multi-year, multi-billion-dollar deal to supply Microsoft with tens of thousands of NVIDIA GB300 GPUs. The 2025–2026 era marks a step-change: a $1.5B structured raise, a gigawatt-scale data-center strategy, and the hiring of telecom veteran Michel Combes as CEO. The bull case is the durable explosion in AI training and inference demand; the bear case is brutal competition from CoreWeave, hyperscalers, and well-funded peers like Crusoe and Nebius, plus thin margins and dependence on NVIDIA supply.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki