Palo Alto Networks Buys Applied Compute for $4.5B

Palo Alto Networks acquisition of Applied Compute to secure open-weight AI models

The Palo Alto Networks acquisition of Applied Compute is a $4.5 billion all-stock deal. The cybersecurity company says it wants to secure open-weight AI models for corporate customers, and it expects the purchase to add more than $1 billion to its Next Generation Security business by fiscal 2029.

Applied Compute runs a platform for post-training and inference of AI models. Palo Alto Networks announced the agreement alongside its fiscal first-quarter 2027 earnings on Wednesday.

Why a security vendor wants a model-training shop

Chairman and CEO Nikesh Arora framed the purchase around two pressures on customers. The first is patching. On the earnings call he said that several major disruptions in recent weeks, including outages at large enterprises and governments, were partly caused by the side effects of patches and configuration changes. In his view, IT teams cannot fix vulnerabilities fast enough without putting production systems at risk.

The second is the rise of AI-assisted cyberattacks. Palo Alto Networks is answering AI-driven risk with more AI, and it is spending heavily to do it.

In its announcement, the company said the future of enterprise AI is hybrid, with a large share of businesses running open-source and open-weight models tuned to their own data, cost and control needs.

What Arora said about open-weight adoption

Arora argued that every enterprise will run a mix of open and closed models, and that each one needs protection. He called Applied Compute’s post-training and inference platform the best the company has seen, and praised its research team as exceptional.

He also pointed to a number. According to Arora, open-weight models have grown from under 5 percent of token traffic to more than 60 percent over the past year. He tied that shift to a wish for sovereignty and control, and to a desire among companies to avoid vendor lock-in and keep hold of their data.

Take the figure as the company’s own claim. It comes from an executive who has a stake in the trend continuing.

The infrastructure bill

The price tag is only part of the commitment. Arora said Palo Alto Networks expects to build infrastructure that supports more than 100 megawatts of inference capacity. He described the spending as a modest capital expenditure commitment of less than $3 billion through fiscal 2030.

So the full outlay includes the stock used to pay for Applied Compute and a separate build-out of inference capacity. For a company best known for firewalls and security software, that moves it a long way toward running AI workloads itself.

A shaky backdrop for open-weight models

The deal lands at an awkward moment for the open-weight camp. In July, Nvidia co-founder Jensen Huang defended open-weight models as a counterweight to the closed commercial models sold by Anthropic, Google, Microsoft and OpenAI. He has since resigned as Nvidia’s CEO, and some in the industry worry that open-weight supporters have lost a powerful voice in Washington.

That matters because the US government is cautious about the spread of models built in China, and open-weight releases from Chinese developers are a large part of that conversation. A security vendor that promises to harden open models for corporate use is selling into that unease.

What the deal means for customers

For enterprise security teams, the practical question is what Palo Alto Networks will actually ship. The company has not detailed products in the announcement. The stated aim is to secure open and open-weight models that companies tune on their own data, which suggests tooling for training pipelines and for the inference stage where those models answer queries.

  • Price: $4.5 billion, paid entirely in stock.
  • Revenue goal: more than $1 billion added to Next Generation Security by fiscal 2029.
  • Infrastructure: over 100 megawatts of inference capacity, with capital spending under $3 billion through fiscal 2030.

What happens next

Deals of this size usually need regulatory review and a formal closing, and the company has not given a closing date in the figures above. Watch for details on how Applied Compute’s platform will be folded into the security product line, and for whether rivals respond with their own purchases of model-training teams.

The bigger test is demand. If open-weight models really do carry most enterprise token traffic, as Arora says, a security company that owns the training and inference layer will have a strong position. If the 60 percent figure proves narrow, the Palo Alto Networks acquisition will look like an expensive bet made at a peak of enthusiasm.

Source: theregister

Written by

Saksham Bhargava

Saksham Bhargava runs Alteroid and has been writing about consumer technology for over five years, covering iPhone, Android, Windows, AI tools and gaming. With an MCA and six years of hands-on web development experience, he tests the features he writes about on his own devices and focuses on clear, practical guides that solve real problems. Away from the screen, he enjoys travelling to places most tourists skip.