Lambda shifts to owning AI data centers

This shift turns Lambda from a renter of GPU capacity into an owner of the bottleneck that matters most, power ready AI data center space. Once Lambda signs multi year contracts with customers like Microsoft, NVIDIA, and frontier labs, owning or tightly controlling the facility lets it plan power, cooling, and cluster layout around those workloads instead of competing for third party capacity every time it needs to expand.1 2 3 4

• The move fits Lambda’s broader jump upmarket. Lambda grew from roughly $425M in annualized revenue at the end of 2024 to $760M by the end of 2025, while dropping lower priority hardware and inference products to focus GPU capacity on large training contracts that run for years, not hours.1 2 3
• Owning and building matters because GPU clouds are now constrained less by finding customers than by finding powered sites. Lambda’s 2026 footprint includes a 24MW Kansas City AI factory, more than 30MW through EdgeConneX in Chicago and Atlanta, and a 21MW long term presence at Prime’s Vernon campus, showing a move from reselling compute toward shaping physical supply.3 4
• This is the same playbook the scaled leaders are using. CoreWeave used long term contracts and asset backed financing to expand its lead to 33 facilities and 250,000 plus GPUs, while Crusoe is building and co owning massive campuses like Abilene. In this market, the winner is increasingly the company that can line up land, power, debt, and customers at the same time.1 4 5

The next phase is a race to turn contract backlog into owned megawatts. As more revenue comes from hyperscalers and frontier labs, Lambda’s valuation and financing capacity will depend less on selling developer friendly GPU instances and more on proving it can build, fill, and operate dense AI facilities as reliably as the biggest infrastructure providers in the market.1 2 4 5