Palo Alto Networks said on November 18, 2026 that it had signed a definitive agreement to acquire Applied Compute, a young company whose platform lets businesses post-train and run open-weight AI models. The cybersecurity company is paying entirely in stock, in a transaction it valued at $4.5 billion, subject to adjustments.
Applied Compute's platform and its AI research team will move to Palo Alto Networks. Yash Patil, the startup's co-founder and chief executive, is set to become Palo Alto Networks' Co-Chief AI Officer once the deal closes. Palo Alto Networks said Applied Compute would keep working with its existing mix of AI-native companies, digital natives and large enterprises, and that it would also offer the platform to its own customers around the world.
The deal requires regulatory clearance and other customary conditions. Palo Alto Networks expects it to close in the third quarter of its 2027 fiscal year. The announcement came on the same day the company was scheduled to report results for its fiscal first quarter of 2027, and it said it would share more details about the purchase on that call.
What Applied Compute built
Applied Compute runs a cloud service that developers use to train large language models and to serve them, according to SiliconANGLE. It also provides a team of machine learning researchers who work on customer projects, for example by building evaluations that measure the quality of newly trained models.
On the training side, the platform relies on a method called asynchronous reinforcement learning. In a typical setup, the part of the system that generates training data and the part that updates the model's weights take turns, each waiting for the other. Applied Compute's approach lets both run at once, which shortens development. The platform also supports self-distillation, a technique that does away with the separate teacher models researchers normally use to guide a training run.
For inference, an autoscaling engine changes the amount of hardware assigned to a model as user traffic rises and falls, and an online reinforcement learning feature periodically retunes a model's weights based on the prompts it receives. The company also sells Shift-1, a router that directs each request from a coding agent to whichever model can handle it at the lowest cost and can divide complex tasks among several models.
In an April 2026 post, the founders described their aim as helping large companies train agents on their own data and workflows, with the systems running inside each customer's own secure environment and under its control.
“Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured. Applied Compute has built the best post-training and inference platform we've seen, and its research team is exceptional. We saw that firsthand as a prospective customer, and our Unit 42 Frontier AI Assessment confirmed their strong security foundations. Together, we will be a leading provider, helping customers train, adapt and deploy open models securely, whether in the public cloud or in their own sovereign environments.”
Why Palo Alto Networks is buying
Palo Alto Networks said it expects enterprise AI to be hybrid, with many large organizations running open-source and open-weight models adjusted for their own data, costs and need for control. The company said no large provider currently offers such models securely and that it intends to fill that gap. Customers would be able to train, adapt and deploy open-weight models with security built in, either in the public cloud or in sovereign environments meant to help with compliance.
The research team is also part of the rationale. Palo Alto Networks said the group would help it add AI across its product range and train its own proprietary large and small models.
Nikesh Arora, Palo Alto Networks' chairman and chief executive, said the company had seen the platform at work while evaluating it as a possible customer, and that an assessment by its Unit 42 team had confirmed that the startup had strong security foundations. Patil said joining Palo Alto Networks would put security at the center of the startup's products and help it win business with very large organizations.
Palo Alto Networks says it has more than 70,000 customers.
The founders and the funding
Applied Compute was founded by Yash Patil, Rhythm Garg and Linden Li. SiliconANGLE reported in early November that the company had launched less than 18 months earlier.
On April 8, 2026, the founders announced an $80 million round at a post-money valuation of $1.3 billion. Kleiner Perkins led it, and Elad Gil, Lux, Greenoaks, Neo and Hanabi took part again. The company said that round brought its total funding to $160 million.
In the first week of November, Applied Compute announced a further $350 million at a $3.25 billion valuation, SiliconANGLE reported. Elad Gil led that round, with Nvidia, Shopify Ventures, HubSpot Ventures and more than a dozen other investors joining. Together, the reported financing comes to about $510 million.
At the time of the November round, the company said its annualized recurring revenue had passed $100 million and that its customers included Microsoft and DoorDash. It said it planned to spend the new money on expanding its market presence and buying more infrastructure. Neither company disclosed Applied Compute's headcount.
From fundraise to sale in two weeks
The purchase price values Applied Compute at roughly 1.4 times the valuation it set in early November and more than three times the figure from April. Palo Alto Networks did not say how many shares it will issue.
In the blog post announcing the November round, the founders wrote that training, inference and routing were only the first parts of a broader set of tools, which they said would eventually include systems to run, monitor and secure AI models. That last item is the core business of their new owner.




