Back

Cybersecurity firm Palo Alto Networks strikes $4.5 billion stock deal to buy Applied Compute, which trains companies' AI

Applied Compute's AC2 dashboard showing a chart of a model's scores across training steps, a table of training runs and a sample answer being graded
Image: Applied Compute

Palo Alto Networks, a large cybersecurity company, has agreed to buy the AI startup Applied Compute in an all-stock deal it values at $4.5 billion. Applied Compute takes AI models whose files anyone can download, gives them extra training on a company's own data, and runs them for that company, and Palo Alto plans to offer that with security built in. The deal still needs regulators' approval, and the company expects it to close between February and April 2027. Its announcement does not say what the combined products will cost or when they will be ready.

Palo Alto Networks, a large cybersecurity company, said on Nov. 18 that it has agreed to buy Applied Compute, an AI startup founded less than two years ago, in a deal paid in Palo Alto shares rather than cash. Palo Alto values the deal at $4.5 billion, a figure that can still change. Sixteen days earlier, on Nov. 2, Applied Compute said investors had valued it at $3.25 billion when they put $350 million into it, so Palo Alto's price is more than a third higher than that value.

Applied Compute works with open-weight models: the models' trained files are published, so anyone can download them and run them on their own computers. Applied Compute gives such a model extra training that adapts it to a specific job, using a company's own data, and then runs the model to answer the company's requests. Applied Compute says companies such as Microsoft, Shopify and DoorDash use it to build and run AI tailored to their work, and that its sales reached a pace of $100 million a year.

Palo Alto says it believes many big businesses will run AI of this kind, tuned to their own data, costs and need for control, and that no large provider offers such models securely today. It plans to let companies train, adapt and run these models with security built in, on rented cloud computers or on systems the customers keep under their own control. "Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured," said Nikesh Arora, Palo Alto's chief executive. He said Palo Alto saw Applied Compute's work firsthand when it was weighing whether to become a customer.

Applied Compute's team will join Palo Alto, which says the team will help train its own AI models. Applied Compute's co-founder and chief executive, Yash Patil, will become Palo Alto's co-chief AI officer once the deal closes. Palo Alto says Applied Compute will keep serving its customers, and that Palo Alto will offer the service to its own customers, which it puts at more than 70,000.

The deal still needs regulators' approval and must meet other usual conditions. Palo Alto expects the deal to close in Palo Alto's fiscal third quarter of 2027, which runs from February to April 2027. Palo Alto says the new services are not on sale yet and may not arrive when expected, or at all.

Palo Alto's announcement does not say how many shares Applied Compute's owners will receive, or what the new products will cost or when they will be ready.