LAMBDA CLOUD

Stephen and Michael Balaban started Lambda in 2012 as an AI photo-editing app. It used computer vision to recognize faces.

The app worked, but the compute bills did not. Training their models on Amazon's cloud cost tens of thousands of dollars.

The brothers figured they could build their own GPU machines cheaper. In 2017 they started building GPU clusters.

In 2018 they opened that cluster to the public as one of the first clouds made just for deep learning. The side project became the whole company.

AI companies need Nvidia GPUs to train and run their models. Nvidia chips are expensive and hard to get.

Lambda buys them in bulk, racks them in data centers, and rents the compute by the hour. It also sells physical GPU workstations and servers to companies that want their own.

So Lambda makes money two ways. It sells the machines.

And it rents time on the machines it keeps. Basically, it is a landlord for AI compute.

Lambda Cloud is the main product — on-demand Nvidia GPU instances you rent by the hour to train or run AI models. Lambda also sells hardware, including the Vector workstations and Hyperplane servers used in offices and labs.

For big customers there are private clusters reserved just for them. The pitch is always the same.

The fastest Nvidia chips, less hassle, better prices than the giants.

Lambda's edge was starting early and staying focused. It only did one thing — GPUs for machine learning — while AWS and Google treated AI as one product among hundreds.

When the AI boom hit in 2023, Lambda already had the hardware, the relationships, and a reputation among researchers. It leaned hard into being the developer-friendly, no-nonsense option.

Then it raised money fast to buy more chips. Revenue jumped around 79% in a single year to roughly $760 million.

Lambda's whole business depends on getting Nvidia chips, and so does everyone else's. It competes with CoreWeave, with cloud giants like AWS and Microsoft, and even with Nvidia's own plans.

Buying billions of dollars of GPUs is brutally capital-heavy. If AI demand cools, Lambda is stuck with warehouses of chips that lose value fast.

It is a great business when compute is scarce. It gets scary the moment compute is not.

Lambda has raised over $2.3 billion. Nvidia itself is an investor, which matters a lot when Nvidia decides who gets chips first.

The February 2024 Series C brought in $320 million at a $1.5 billion valuation. A $480 million round followed in early 2025.

Then in November 2025 Lambda raised a $1.5 billion Series E led by TWG Global. The money all goes to the same place — more GPUs.