Palo Alto Networks Expands into Enterprise AI with $4.5B Deal
Enterprise security outfit Palo Alto Networks snaps up a startup with a platform for post-training and inference of open-weight AI models.
Palo Alto Networks, a cloud company that repositioned itself as an integrated cybersecurity provider, has ventured into AI infrastructure by acquiring a startup with a platform for post-training and inference of open-weight AI models. Applied Compute, co-founded in early 2025 by three former OpenAI researchers, grew from zero revenue to over $100 million in annual revenue in under 18 months while working with major AI companies.
Palo Alto Networks has announced it will acquire Applied Compute in an all-stock transaction valued at $4.5 billion. This acquisition caught the market by surprise not only because of its price tag, but also because Applied Compute operates outside the traditional cybersecurity sector.
First and foremost, the deal reflects a broader strategic shift in which major technology companies are now seeking to develop their own AI capabilities and gain greater control over AI systems, rather than relying solely on models from companies such as OpenAI and Anthropic. Applied Compute specializes in post-training and inference infrastructure for open-weight AI models, which enables organizations to customize and deploy AI systems for specific applications.

Palo Alto Networks gets AI
Applied Compute helps companies adapt and run open-weight models, AI systems whose trained settings are publicly released. Apparently, they present security challenges, including the risk of model tampering, data leakage, and misuse by bad actors. Post-training data is usually a company’s most confidential material—tickets, contracts, and customer records—and the adapted weights can retain fragments of it.
As a result, the model can inherit its training-set classification, so it can be traced and retired when a source is withdrawn. “Open-weight model implementations will allow us to deliver secure, open-weight models to our customers to enable adoption of secure AI in the enterprise,” said Nikesh Arora, Palo Alto’s chairman and CEO.
Catharine Trebnick, senior research analyst at Rosenblatt Securities, told Reuters that AI tools are surfacing vulnerabilities faster than companies can patch them, pushing companies to consolidate onto fewer platforms. This deal, for instance, could allow Palo Alto Networks to develop cybersecurity-specialized models.
As the company moves into enterprise AI infrastructure with the acquisition of CyberArk, expected to close in the third quarter of 2027, it could embed security directly where enterprises train, govern, and run custom open-weight models. The question is why didn’t Palo Alto Networks build this technology with in-house engineering talent? A simple answer: Palo Alto Networks has engineers, but it doesn’t have time.

AI models embedding security
The deal signals a deeper integration of advanced AI capabilities into mainstream security products. It comes as a heightened threat environment pushes organizations toward broader security solutions. “Applied Compute would now add the models themselves to Palo Alto Networks, already at the center of many companies’ firewalls, chips, and identity controls,” said Sanchit Vir Gogia, chief analyst at Greyhound Research.
This acquisition could further spur consolidation in the cybersecurity AI space at a time when companies bolster their AI offerings. Especially when AI talent is still scarce within large security firms; it largely sits in small startup firms. Palo Alto Networks evaluated nine providers before settling on Applied Compute.
The acquisition also marks a shift in Palo Alto Networks’ strategy as AI reshapes both the cybersecurity market and enterprise technology’s infrastructure requirements. Palo Alto Networks is positioning itself not only to secure the growing use of AI models and autonomous agents, but also to provide some of the technology required to train, adapt, and run those models.
In other words, the deal will put Palo Alto Networks in a position to train, adapt, and deploy open-weight AI securely in both public cloud and sovereign cloud environments. As Gogia noted, while Palo Alto Networks has sovereignty credentials for security data, including customer-held keys, none of that yet extends to AI accelerators.

Eventually, Palo Alto Networks aims to offer a technology platform that helps organizations securely train and deploy open-weight models as they expand their use of AI. This is especially true when no one is currently offering secure open-weight models. TD Cowen analyst Shaul Eyal summed it up by saying Applied Compute extends Palo Alto’s push beyond traditional cybersecurity into enterprise AI.
The $4.5 billion transaction is subject to regulatory approval and other customary conditions and is expected to close during the third quarter of 2027. Once the transaction is complete, Yash Patil, co-founder and CEO of Applied Compute, will become co-chief AI officer at Palo Alto Networks.
Also read:
When AI Meets Hardware Security
Open-Weight Models and the Future of AI
AI Is Accelerating Cyberattacks. Is Your Hardware Ready?
SambaNova Abandons Intel Acquisition, Raises Funding Instead
RELATED TOPICS: AI INFERENCE, AI INFRASTRUCTURE, AI TRAINING, CLOUD COMPUTING, CYBERSECURITY, MERGERS & ACQUISITIONS, SEMICONDUCTORS
COMPANIES: APPLIED COMPUTE, PALO ALTO NETWORKS
Majeed Ahmad, Editor in Chief of EDN and Planet Analog, has covered electronics design industry for longer than two decades. During this period, he has worked in various editorial positions, including assignments for EE Times Asia and Electronic Products. He holds a Masters’ degree in telecommunication engineering from Eindhoven University of Technology. Follow Majeed on LinkedIn
Source: EE Times




