Palo Alto Networks to Buy Yash Patil's Applied Compute for $4.5 Billion

8 Min Read
Palo Alto Networks to Buy Yash Patil's Applied Compute for $4.5 Billion

A security giant moves into building AI models

Palo Alto Networks has agreed to buy Applied Compute, a startup founded less than 18 months ago, in a deal valued at $4.5 billion. The transaction, announced on Nov. 18, is to be paid for in Palo Alto Networks shares. Applied Compute co-founder and chief executive Yash Patil, 23, will become the cybersecurity company's joint chief AI officer once the deal closes.

Contents
  1. A security giant moves into building AI models
  2. What does Applied Compute make?
  3. Why is a cybersecurity company buying an AI platform?
  4. What will Patil do at Palo Alto Networks?
  5. How large could the business become?
  6. The deal arrives during a strong quarter
  7. What could customers gain, and what needs attention?
  8. Key dates and figures
  9. Key Points

The proposed purchase brings Palo Alto Networks into the business of helping companies train and run artificial intelligence models, as well as protecting them. Applied Compute's software lets customers adapt open weight models using their own information and run them in environments they choose. The companies say the platform will serve customers through public cloud services and in sovereign environments, where organisations seek greater control over where their data is stored and processed.

The agreement still needs regulatory clearance and other customary closing conditions. Palo Alto Networks expects it to close in its fiscal third quarter of 2027, which runs from February through April and ends April 30.

What does Applied Compute make?

AI models learn patterns from data during training. An open weight model makes the numerical settings that shape its behaviour available for others to use and adapt. A company can then tune a model with its own data, such as product records or internal instructions, rather than relying only on a general purpose system operated by another provider.

Applied Compute's platform supports work after a model's initial training, including adjusting it for a particular task and running it to answer users' requests. The second process is called inference. Applied Compute says its software can help companies train, serve and improve models built for their own needs.

The startup was founded by Patil, Rhythm Garg and Linden Li. It counts Microsoft and DoorDash among its customers, according to company details reported around the announcement. Earlier in November, it raised $350 million at a $3.25 billion valuation. The $4.5 billion purchase price is about 38% above that valuation, although the acquisition value is subject to adjustments.

Advertisement

Why is a cybersecurity company buying an AI platform?

Palo Alto Networks chairman and chief executive Nikesh Arora says businesses will use a mix of open and closed models, and that both kinds need protection. He described the aim as helping customers train, adapt and deploy open models securely, whether they use public cloud services or their own controlled computing environments.

Arora said Palo Alto Networks first assessed nine providers before deciding to pursue Applied Compute. He said its research team and technical work stood out in the area of adapting open weight models. Applied Compute's researchers are expected to help build AI across Palo Alto Networks' products and train the company's own large and small models.

Arora has also argued that security tools need to understand how a particular AI application is meant to behave. A general security filter may spot familiar risks, but it may not know whether a particular action fits the rules of a company's application. Palo Alto Networks' approach is to train a supervisory AI model alongside an application, so one system can monitor another.

Arora explained why the company believes it needs to be involved in model training as well as security.

We have to get in the training business to be able to build a good supervision business and that's our bet.

The idea is to make security part of the process of building and running an AI application. Palo Alto Networks says its customers will be able to use the acquired platform for open weight models while continuing to choose closed models from other providers.

Advertisement

What will Patil do at Palo Alto Networks?

Patil is Applied Compute's co-founder and chief executive. He will take on the joint chief AI officer role after the acquisition closes, and the startup's team is expected to join Palo Alto Networks. The original account describes Patil as an Indian American Stanford dropout.

Patil said the startup was built to make it easier for companies to train, run and improve models of their own. He said joining a cybersecurity company would put security at the centre of its products and help the team reach larger organisations.

Joining Palo Alto Networks will put security at the center of all our products and accelerate our traction with the world's largest organizations.

The companies say Applied Compute will continue serving companies built around AI, digital businesses and large enterprises. Palo Alto Networks plans to take the platform to its global customer base. The startup's team is also expected to contribute to Palo Alto Networks' wider AI research and product work.

How large could the business become?

Applied Compute reached more than $100 million in annualised revenue within 18 months of its founding, according to Palo Alto Networks executives. The cybersecurity company expects the acquired business to contribute more than $1 billion in annual recurring revenue by fiscal 2029. Annual recurring revenue estimates the yearly value of ongoing customer subscriptions.

Palo Alto Networks also plans to build more than 100 megawatts of computing capacity over the next three to five years. Chief financial officer Dipak Golechha said most of the planned spending would go to graphics processing units, or GPUs, and servers, with a smaller amount for facilities. GPUs are specialised chips widely used to train and run AI models. The company has said capital spending for this effort will remain below $3 billion through fiscal 2030.

Arora said open weight models had grown from less than 5% to more than 60% of AI token traffic over the past year. Tokens are small pieces of text processed by an AI model. He did not explain how the figures were calculated. His broader point was that businesses may spend more on tools that secure models as their use of AI grows.

Advertisement

The deal arrives during a strong quarter

Palo Alto Networks announced the acquisition alongside results for its fiscal first quarter, which ended Oct. 31. Revenue rose 39% from a year earlier to $3.44 billion. Net income was $482 million, compared with $334 million a year earlier. The company also reported non-GAAP earnings of $1.05 per share, a measure that excludes some costs and adjustments.

The company raised its fiscal 2027 revenue forecast to between $14.52 billion and $14.57 billion, above the average analyst estimate of $14.16 billion cited at the time. Palo Alto Networks shares rose after the market closed. Market accounts put the gain at 8.5% and about 10%.

The acquisition follows other major purchases as Palo Alto Networks expands its security business. Its 2026 acquisitions included identity security company CyberArk and observability firm Chronosphere. The Applied Compute deal adds an AI model training platform and research staff to that expansion.

What could customers gain, and what needs attention?

For customers, the stated benefit is the option to adapt models to company data and choose where to run them, while buying security services from the same provider. Sovereign environments are designed to give organisations more control over data location and operating arrangements, which can matter for public bodies and companies facing rules about sensitive information.

That combination also means customers should pay attention to who controls trained models and the data used to build them. Sanchit Vir Gogia, chief analyst at Greyhound Research, said companies should seek clear terms covering export of trained model weights, consent before customer data is used for training, the right to leave after a change of ownership, and protection against price changes. He also said a separate party should test an AI system's safety.

Gogia described the risk of relying on a single team to build and supervise an AI application.

A supervisor trained by the same hands as the system it supervises is a colleague, not a check.

Regulatory clearance remains a condition of the purchase, and the companies have not announced a final closing date beyond the expected fiscal third quarter. Palo Alto Networks has also not specified a region list or operating date for its planned dedicated computing capacity. Those details will matter to customers deciding where sensitive models and data can be hosted.

Key dates and figures

  • Nov. 18, 2026: Palo Alto Networks announced the $4.5 billion proposed acquisition.

  • Earlier in November: Applied Compute raised $350 million at a $3.25 billion valuation.

  • February to April 2027: Expected closing period, subject to regulatory clearance and other conditions.

  • Fiscal 2029: Palo Alto Networks expects Applied Compute to contribute more than $1 billion in annual recurring revenue.

  • Within three to five years: The company aims to build more than 100 megawatts of computing capacity.

Key Points

  • Palo Alto Networks agreed to acquire Applied Compute for $4.5 billion in a share based transaction.

  • Yash Patil, 23, will become the company's joint chief AI officer after closing.

  • Applied Compute makes software for adapting and running open weight AI models using company data.

  • The transaction requires regulatory clearance and is expected to close in Palo Alto Networks' fiscal third quarter of 2027.

  • Palo Alto Networks forecasts more than $1 billion in annual recurring revenue from Applied Compute by fiscal 2029.

Share This Article

You May also Like