Palo Alto Networks aims to secure AI models with Applied Compute
Palo Alto Networks has reached an agreement to acquire Applied Compute in an all-stock deal valued at $4.5 billion. The goal: securing the phase where companies adapt and deploy their AI models.
Palo Alto Networks announced on November 18, 2026, a definitive agreement to acquire Applied Compute, a startup specializing in adapting and running open-weight AI models in enterprises. The all-stock transaction is valued at $4.5 billion, subject to adjustments and regulatory approvals. It aims to strengthen the cybersecurity group’s strategy in a market where AI models are no longer just to be protected but also to be operated smoothly. Not exactly a minor detail.
Securing AI as it goes into production
The core thesis of Palo Alto Networks is simple: companies will mix open and closed models, and each must be secured. Nikesh Arora, the group’s CEO, summarized this logic by saying: “Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured.” The acquisition therefore targets the most sensitive layer, where a model is adapted, served, and integrated into business workflows, not just tested in a lab.
Applied Compute, which will continue to serve its current clients if the deal closes, works on use cases of customization, inference, and deployment in public or sovereign environments. For Palo Alto Networks, the interest goes beyond classic cybersecurity: the group also wants to integrate this component into its product portfolio and train its own large and small models. Yash Patil, co-founder and CEO of Applied Compute, is set to become co-head of AI at Palo Alto Networks after closing.
A young startup, but already heavily computing-intensive
Applied Compute published several figures on November 2, 2026, shedding light on the appeal of the deal. The company claims to have reached an annualized revenue of $100 million in less than 18 months after its founding. It also raised $350 million at a valuation of $3.25 billion in a round led by Elad Gil. Annualized revenue measures the current revenue run rate over twelve months, without corresponding to a full fiscal year.
In terms of usage, the startup states that its AC Inference service processed more than 1 trillion tokens per day. Since July 2026, clients had also launched more than 1,700 trainings on AC2, its Agent Cloud platform. Finally, 40% of clients were using both training and inference at the time of the announcement. This mix of activity is precisely what Palo Alto Networks is looking for: a tool that does not just monitor AI but supports its production deployment.
What Palo Alto Networks is really buying
At its core, Palo Alto Networks is not just acquiring a team but a research and execution expertise. Applied Compute indicated that its researchers and engineers represented 85% of the workforce, with total headcount having tripled since January 2026. This technical density matters because it can feed the group’s projects on proprietary models, a still rare area among cybersecurity players.
The commercial interest is broader. Palo Alto Networks has already said it studied Applied Compute as a potential client, and its security assessment conducted by Unit 42 would have confirmed its foundations [to be verified]. If the merger materializes, the group will move from a posture of protecting AI uses to a deeper presence in the value chain, where quality, cost, and compliance of models are decided. It is also where inference bills start to climb without asking permission.
Key points
- Acquisition announced on November 18, 2026.
- All-stock transaction valued at $4.5 billion.
- Applied Compute targeted $100 million in annualized revenue.
- 1,700 trainings launched on AC2 since July 2026.
- Yash Patil to join Palo Alto Networks after closing.
- Closing expected in fiscal third quarter 2027.
By the numbers
- $4.5 billion — value of the deal announced by Palo Alto Networks, November 18, 2026.
- $100 million — annualized revenue reported by Applied Compute, November 2, 2026.
- $350 million — funding announced by Applied Compute, November 2, 2026.
- 1 trillion tokens per day — volume claimed for AC Inference, November 2, 2026.
- 85% of workforce — share of researchers and engineers at Applied Compute, November 2, 2026.