Latest

AI-generated illustration of a shield standing in front of a rack of servers. It is not a picture of anything in this story. Illustration: Grok.

By news editor

Palo Alto Networks has agreed to acquire Applied Compute, a platform for post-training and inference of open-weight AI models, in an all-stock transaction the company values at $4.5 billion, subject to adjustments. The announcement takes a cybersecurity vendor into the business of training, adapting and deploying AI models for its customers.

Both sides have signed a definitive agreement. Closing is expected in Palo Alto Networks' fiscal third quarter of 2027, subject to customary conditions that include regulatory clearance. The release does not say how many shares will be issued, and it gives no revenue, customer count or headcount for Applied Compute.

For scale, the deal arrived with the buyer's fiscal first quarter results. Total revenue for the quarter ended October 31, 2026 grew 39% year over year to $3.44 billion, and Next-Generation Security annual recurring revenue grew 66% to $9.73 billion. The transaction value is larger than that quarter's total revenue.

Open-weight models are AI models whose trained parameters are available for a company to run and adapt on infrastructure it controls. Post-training is the work of adapting such a model after its initial training. Palo Alto Networks says it believes a large share of enterprises will run on open-source and open-weight models tuned to their own data, cost and control needs, in public clouds or in sovereign environments.

Its release then makes a competitive claim: "Today, no large provider is delivering those models securely." The statement names no provider and cites no evidence.

Chairman and CEO Nikesh Arora said in the release: "Every enterprise is going to run a mix of open and closed models, and every one of those models needs to be secured." He added that Palo Alto Networks had looked at Applied Compute as a prospective customer and that an assessment by its Unit 42 team confirmed the target's security foundations.

People are part of the purchase. Applied Compute's team will join the buyer, and co-founder and CEO Yash Patil will become Co-Chief AI Officer after the close. Palo Alto Networks says the research team will help train its own large and small AI models and embed AI across its products. Applied Compute will continue to serve AI-native companies, digital natives and large enterprises, according to the release.

Running models requires hardware, and that part of the plan is not in either release. Network World, in an article it says first appeared on CSO, reports that CFO Dipak Golechha told the earnings call the company plans to build its own compute capacity, with spending going mostly to GPUs and servers, while continuing to source from cloud partners. Arora said the aim is more than 100 megawatts of compute in the next three to five years, according to the same report. Neither release gives a spending figure for that build-out.

One analyst quoted in that article questions the design. Sanchit Vir Gogia, chief analyst at Greyhound Research, argued that a supervisory model trained by the same vendor as the system it oversees is not an independent check, and that the sovereign hosting offer so far has no named region, operator or date. He advised Applied Compute customers to use the months before closing to secure ownership of their trained weights in an exportable format and a right to exit on a change of control.

The next thing to watch is the closing, which would fall between February and April 2027 on the fiscal calendar Network World describes, if regulators clear the deal. Before then, Palo Alto Networks has yet to disclose the shares it will issue, Applied Compute's financials, or the regions and dates for the sovereign hosting it describes.

Leave a Reply