Nikesh Arora puts the security equipment already installed and too old to cope with AI at $1 trillion, and says the AI data centers being built on top of it will need a security stack that does not exist yet.
Nine months ago he was telling people his industry had been written off as a casualty of artificial intelligence. Palo Alto Networks then reported a quarter that beat on almost every line, with guidance for the current quarter and the full year to match.
"It's q4, it's party time, Jim."
Arora is chairman and chief executive of Palo Alto Networks, a long-time holding of Cramer's charitable trust, and he came on straight from the earnings call.
I listened to the full segment so you can skip it.
Here are the 7 takeaways that matter.
👤 Guest: Nikesh Arora, chairman and chief executive of Palo Alto Networks, who spent ten years at Google before taking the job
🎙️ Host: Jim Cramer, who hosts Mad Money and runs the charitable trust behind the CNBC Investing Club, which has held Palo Alto Networks for years
📰 Published: 1 September 2026 on CNBC (Mad Money w/ Jim Cramer)
🟢 Spotify | 🟣 Apple Podcasts | 🔗 CNBC | ⏱️ 10 min
Key Takeaways
The industry that was supposed to be destroyed by AI is being bought because of it Arora says nine months ago he was "guilty and convicted of near death"; the quarter says otherwise
$1 trillion of security equipment is old enough that it cannot cope with an attack running at machine speed His arithmetic: a seven-year average life against $200 billion to $300 billion of annual spending
The AI data-center build-out needs a security stack nobody has built yet He puts the capital spending behind it at $5 trillion over five years
One rival's model launch did what Arora could not do in eight years of telling customers they were exposed About 2,000 companies have since had a conversation with Palo Alto about it
Governments are better prepared than companies, not worse
Almost no customer is ready to deploy a kill switch for its own agents What they want instead is the ability to identify an agent and stop it mid-task
The platform strategy the market punished three years ago is the reason this quarter worked
1. Q4 doubled the platforms
Cramer set the scene on the sector rather than the company. "The cybersecurity stocks have been climbing steadily higher for months now, ever since Wall Street figured out that this industry wouldn't be made obsolete by artificial intelligence." He said Palo Alto had more than doubled from its April lows to its highs a few weeks earlier, had pulled back into the print, and had then beaten on almost every key line with strong guidance for both the quarter and the year.
Cramer read the platform count off his own card rather than getting it from Arora. "This stat roughly 220 new net new platforms in the fourth quarter. Double the 110 from the last quarter not year over year but the last quarter."
Arora's answer was a quarter-end joke rather than an explanation. "It's q4, it's party time, Jim."
He said the early signs are customers wanting to adopt platforms, and customers wanting a head start on what AI does to attacks
2. From near death to a feast
Arora restated the position he took at the start of the year and then abandoned it. "Look, nine months ago, I said we were guilty and convicted of near death because AI was going to eat our lunch, breakfast and dinner. It seems like that's not the case. Seems like we're going to have to have the feast with them."
He dates the change to one product launch rather than to anything his own company did. "And the Mythos moment has just become a net new beginning for the cybersecurity industry, because the world has realized that we have to pay attention to cyber because AI is going to be weaponized by bad actors, and they will need the large players in the cybersecurity space to help protect them." The claim inside that is a market-structure one: that buyers dealing with weaponized AI will consolidate on the biggest vendors
3. The $1T of legacy security
Cramer put Arora's own framing back to him and translated it: modernizing $1 trillion of legacy security means $1 trillion worth of security that can be hacked.
Arora derived the number rather than asserting it. "Well, think about it. If the average life is seven years and you're spending 2 to $300 billion a year, you've got $1 trillion of security infrastructure that's out there. Nothing that was deployed 7 or 10 years ago is prepared or ready to handle AI at machine speed."
The conclusion he draws is architectural, not a product upgrade. "So you have to make sure that you have to rethink your cyber architecture."
On top of the installed base is the equipment not yet built. He said, "not to miss the fact that on top of that, you're going to see $5 trillion of capex spend in the next five years with people building AI data centers and having tons and tons of agents running around." "You also have to build a net new security stack for that."
4. One minute to resolution
Cramer asked what Palo Alto is actually doing about the agents now running loose in corporate networks.
Arora's answer was that the company built for this before it had to. He said Palo Alto has been consistent about building an AI-first product, that it shipped a new security-operations product, and that it bought an observability company. "We bought a company called Chronosphere, which focuses on AI driven observability." He traced the conviction to his old employer: it was the one thing he brought with him from Google, where he worked ten years ago, that AI was going to be big — though he said he did not know when or how big
The performance claim is a time to fix, not a detection rate. "We've been able to get ourselves to one minute of resolution. We think we can get our customers there."
He would not accept the framing that everyone is exposed. "The good news is not all customers are unprepared." Some, he said, have already deployed the new security-operations capability
5. Governments are readier
Cramer asked, half as a challenge, whether the government is ready.
Arora said governments come out of this better than companies do, and gave two reasons. "Well, governments have a phenomenal property because a lot of them are not cloud first, because they've been very particular about air gapped systems, about sovereign systems. To be fair, they actually have done a much better job." The second reason is people: he pointed to agencies such as the National Security Agency and their equivalents elsewhere, which he said have deep knowledge on both the defensive and the offensive side and have seen the worst attempts in the world
His verdict was qualified rather than reassuring. "So I think they are they are aware, they are prepared."
He added that governments put a far larger share of their information-technology budgets into cybersecurity than companies do, which he said is not a well known fact
6. 2,000 calls after Mythos
Cramer asked what kind of companies are calling about the AI defense program Palo Alto had announced.
The pipeline is a conversation count, not a revenue number. "Jim, we've had about 2000 companies that we've had a conversation with."
What Palo Alto is selling into those calls is a testing service. Arora said the company is the first certified commercial partner for Mythos, so it can bring the model in with its own people and test how vulnerable a customer is, and that it can run the same exercise with another laboratory's model The offer runs on from there to a plan to fix what the tests find and a plan to modernize the infrastructure
The claim he draws from it is about the industry's growth rate, not his own quarter. "So that's 2000 conversations. Not everything's going to happen next quarter, but all I say is this changes the long term growth rate and duration of cyber security, not just for palo alto, but as an industry, because suddenly people are paying attention that you cannot deploy AI successfully if you don't get cybersecurity."
He credits the shift to somebody else's launch event. "And I'll tell you finally, Jim, I've been trying for eight years to tell customers they're not ready. And dario did it in one event just by launching Mythos." "He made every CEO aware that AI has capabilities which we have not all ever thought about."
7. Nobody has a kill switch
Cramer asked how many companies are ready with a kill switch to stop the agents.
Arora expects agent security to be a new category rather than a feature. "I think the agent security space is going to be a net new, amazing space."
On readiness he was blunt. "I think that's going to keep evolving, but customers are not fully ready to deploy the kill switch."
The reason customers are asking is a specific incident. "And we all read about things like OpenAI models escaping and hacking, Hugging Face. That puts a bit of a fear in the minds of our customers saying, I don't want rogue agents in my infrastructure."
What they want is narrower than a kill switch and more useful. "So even if you don't have a kill switch, you definitely need the ability to intercept agents and stop their activity."
Cramer raised the platform strategy analysts had once questioned, and called it the best thing the company has done. Arora agreed and said it cost him at the time. "Well, three years ago, we took a few punches because we said, we're going to go towards the notion of platform" He said the same consolidation already happened in customer-relationship management software, at Workday and in enterprise resource planning "When multiple systems are fragmented, they cannot talk to each other at machine speed."
Bonus Insights
On what the last months have been like, Arora described a reversal of who calls whom. "My life has been great. You know, it's much better when customers call you than you have to go chasing them."
He compared the corporate rush into AI to the last one he lived through. "I remember the 1999, we had the internet race, and a lot of companies were left behind because the internet created a whole new categories of companies. I think this time we're all resolute not to become the has beens, and that's causing CEOs to get more assertive, understand AI, deploy AI." The commercial consequence he draws is that the first question a newly assertive chief executive asks is whether the AI will be secure
Arora also named a recently acquired team working on AI-first security and IT operations, which he said will help make Palo Alto's own products AI-native as well as its customers' infrastructure
Cramer's sign-off disclosed his position again. "My trust is a happy shareholder because we've got a happy Nikesh because this was a monster quarter."
Cramer's own market analysis, his case for a much larger Nvidia buyback, and the interviews with Dell Technologies and CrowdStrike from the same programme are written up separately
Arora's bottom line is that AI turned out to be a demand event for cybersecurity rather than a threat to it, because a trillion dollars of installed equipment and five trillion dollars of new data centers both have to be re-secured against attacks that move faster than people do.
Products, Companies & Tools Mentioned
Palo Alto Networks (Arora's company, which reported a quarter he says was driven by customers consolidating onto platforms and by the new urgency around AI)
Chronosphere (The AI-driven observability company Palo Alto bought, named by Arora as part of building an AI-first security and IT operations business)
Mythos (The frontier model whose launch Arora credits with making every chief executive aware of what AI can do; Palo Alto says it is the first certified commercial partner for it)
OpenAI and Hugging Face (The escape-and-hack incident Arora says put the fear of rogue agents into his customers)
Workday (One of his examples of an enterprise software category that consolidated onto a platform, as he says security now must)
Google (Where Arora worked ten years ago, and where he says he formed the conviction that AI would be big)
National Security Agency (His example of the deep offensive and defensive capability that makes governments better prepared than companies)
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: