Palo Alto drops $4.5B on the open-weight AI it cannot yet secure
Nikesh Arora's all-stock purchase of Applied Compute bets that enterprises will run open models at scale, then pay someone to lock them down.

Palo Alto Networks, led by chairman and CEO Nikesh Arora, has agreed to acquire AI model training platform Applied Compute in a $4.5 billion all-stock deal, announced alongside its fiscal Q1 2027 earnings call. The bet is that open-weight and open-source models become the default enterprise stack, which turns securing hybrid model fleets into the next large security budget line.
Palo Alto Networks is buying its way into the model-training business. The security vendor has agreed to acquire Applied Compute, an AI model training platform, in a $4.5 billion all-stock transaction announced alongside its fiscal Q1 2027 earnings call on Wednesday. Chairman and CEO Nikesh Arora did not undersell the target: "Applied Compute has built the best post-training and inference platform we've seen, and its research team is exceptional," he said in a statement. Palo Alto expects the deal to add more than $1 billion to its Next Generation Security business by FY 2029. That is the headline number, and it is a big one for a purchase of a company most enterprise buyers have never procured from directly.
The reasoning behind the check arrived on the same earnings call. "We believe many of the major disruptions of the last few weeks, including outages across major enterprises and governments, were partially due to unintended consequences of patching and configuration changes, as IT teams simply cannot patch these vulnerabilities fast enough without compromising their production environments," Arora told listeners. "The second effect is the rise of AI-assisted cyberattacks." The Register's read on the strategy is blunt and hard to argue with: in the spirit of hair of the dog that bit me, Palo Alto is ingesting an AI training platform to make open-weight AI more secure for corporate customers. The answer to insecurity brought on by AI is buying security with AI.
The number that actually sold this deal is a traffic stat. "We're already seeing the desire for sovereignty and control compel enterprises to adopt open-weight and open-source models," Arora said. "Early signs of that shift are clear, with open-weight models growing from under 5 percent of token traffic to more than 60 percent over the past year." For anyone outside an AI platform team, token traffic is simply the volume of requests flowing through a model - the closest thing the industry has to an odometer. Palo Alto's claim is that the odometer has flipped from a rounding error to a majority of enterprise work in roughly twelve months. Treat that as the company's own framing rather than an audited market figure, but the direction of travel is the entire thesis behind the acquisition price.
The capital picture is worth sitting with. Palo Alto is paying in stock, not cash, which means existing shareholders absorb the dilution rather than the balance sheet taking a hit. Management expects the deal to contribute more than $1 billion to Next Generation Security by FY 2029, and it is separately preparing to spend real money on the plumbing underneath it. Arora said the company expects to build infrastructure supporting more than 100 megawatts of inference capacity, describing it as a "modest CapEx commitment of less than $3 billion through fiscal 2030." Colocating that much inference capacity moves a software security vendor closer to being an infrastructure operator, with the depreciation schedules, utilization risk, and energy exposure that come with the territory.
The wider political backdrop makes the timing sharper. In its deal announcement, Palo Alto declares that "the future of enterprise AI is hybrid, and that a large share of enterprises will run on open-source and open-weight models tuned to their own data, cost, and control needs." That view has been clouded, as the report puts it, by Nvidia co-founder Jensen Huang's fall from grace. Back in July, Huang came to the defense of open-weight models as a defense against the hegemony of closed-weight commercial purveyors like Anthropic, Google, Microsoft, and OpenAI. His resignation as CEO of Nvidia has led to worries that the open-weight ecosystem has lost some lobbying power at a time when the United States is wary about the uptake of models made in China. Palo Alto is, in effect, placing a multi-billion dollar wager that the ecosystem survives that political chill.
For the executives and boards reading this, the second-order effects are the point. If a majority of enterprise token traffic genuinely moves to open-weight and open-source models, the security perimeter shifts from endpoints, email, and cloud workloads to the post-training and inference layers where those models are tuned and served. That change ripples through procurement: buyers who spent the last decade arguing about vendor lock-in now have to ask who secures a model they downloaded, fine-tuned on their own data, and are running inside their own environment. Sovereignty and control, the two words Arora used to explain the shift, are now procurement arguments with real budget attached, not just compliance talking points.
What to watch next is execution, not announcement. The $4.5 billion price is all stock, so the return shows up first as dilution and only later as the promised FY 2029 revenue. The inference build-out carries a sub-$3 billion capex tag through fiscal 2030, a number that stays modest only if demand for serving those models materializes as advertised. And the open-weight ecosystem still has to navigate a Washington that is nervous about where models are built. Palo Alto has bought the training layer at the moment the market moved. The open question, for peers in every security and infrastructure seat, is whether the securing layer is the next $4.5 billion category, or whether it gets bundled into the platform they already own.
This story's Key Insights and Take-aways are locked.
Create a free account to unlock Executive Actions for one credit.
Register to UnlockAlways free for Executives Club members. Join the Club
More in Business
S&P 500 Enters Correction, Down 10% From Its High on AI Worries
The index's slide past the 10 percent line is a symbolic verdict on investor mood, and the AI trade that powered the rally is now the thing testing it.
Dow and S&P 500 enter correction as Nasdaq logs worst week since 2020
All 11 S&P 500 sectors fell more than 1% in the same week for the first time since April 2025, a breadth signal that matters more than any single index level.
Chip stocks lose $2.3 trillion in a week, worst run in 25 years
A $2.3 trillion weekly wipeout in semiconductors is forcing every board with an AI story to ask whether it owns a thesis or just a crowded trade.



