AI agent traffic reaches 9 times the level of a year ago! Palo Alto Networks (PANW.US) earnings ignite Wall Street's bullish bets on AI security

date
21:38 19/11/2026
avatar
GMT Eight
Wall Street financial giants have broadly issued bullish research reports on the company, emphasizing that cutting-edge AI technology not only substantially expands the underlying models, databases, and agent operating modes that enterprises need to rigorously protect, but also makes separately procured security tools harder to coordinate, driving enterprises to concentrate more budget on platforms capable of unified management of networks, identities, and security operations.
Title context: AI agent traffic reaches 9 times the level of a year ago! Palo Alto Networks (PANW.US) earnings ignite Wall Street's bullish bets on AI security Text: After cybersecurity super-giant Palo Alto Networks (PANW.US) reported earnings, its stock price surged more than 10% at one point in U.S. premarket trading. Wall Street financial giants broadly issued bullish research reports on the company, emphasizing that cutting-edge AI technology has not only greatly increased the underlying models, databases, and agent operating modes that enterprises need to rigorously protect, but also made separately procured security tools harder to coordinate, pushing enterprises to concentrate more budget on platforms that can uniformly manage networks, identities, and security operations. Palo Alto announced strong quarterly results and guidance, and announced the US$4.5 billion acquisition of startup Applied Compute. Wall Street analysts covering the company's target price and rating subsequently became more bullish on the company. Morgan Stanley said Palo Alto's first-quarter results highlighted "growing enterprise demand for cybersecurity and the willingness of enterprises to centralize related spending as environments become harder to manage." Morgan Stanley maintained its "Overweight" rating on the stock. The firm said that although the valuation is not cheap, the company's "long-term importance and its ability to consistently deliver upside surprises make us firmly maintain our Overweight rating." Susquehanna believes that, thanks to the advancement of its platformization strategy and growing market demand for AI security products, the company's latest quarter "significantly exceeded expectations and raised guidance." The firm maintained its "Positive" rating and US$425 target price. The firm said: "We continue to believe that, given its favorable position, Palo Alto is expected to be one of the main beneficiaries of cybersecurity budget growth, industry consolidation, and the platformization trend." Wedbush also maintained its "Outperform" rating and US$400 target price, and continued to include Palo Alto on its "Best Ideas" list. The firm said that as customers increasingly want to use cybersecurity solutions to address continuously evolving threats under the influence of AI, Palo Alto "significantly exceeded expectations across all metrics" and raised its fiscal 2027 guidance. Wedbush said: "Palo Alto delivered a solid quarter and continues to capture growing market demand for its AI solutions." The firm also expects this demand to remain resilient in the coming years. Analysts said the US$4.5 billion acquisition of Applied Compute will give Palo Alto the technology to train and run open-weight AI models. CEO Nikesh Arora said the acquisition will help the company "provide customers with secure open-weight models and accelerate enterprise adoption of AI." Morgan Stanley said the deal is strategically aligned with Palo Alto's efforts to safeguard enterprises' secure adoption of AI, but also noted that it will make Palo Alto's business more capital-intensive. The firm noted that Palo Alto plans to deploy more than 100 megawatts of computing capacity globally over the next three to five years, with cumulative capital expenditure of about US$3 billion through fiscal 2030. Morgan Stanley said: "Although the deal is strategically aligned with the goal of safeguarding enterprises' secure adoption of AI, it also changes the investment characteristics of this business." Palo Alto has already turned cybersecurity demand in the AI era into a revenue driver On November 19, before the U.S. market opened, Palo Alto rose more than 10% at one point, while CrowdStrike, Zscaler, and Fortinet rose 5% to 7% in tandem, and Nasdaq index futures rose nearly 2% over the same period. There was an important basis for this round of buying: Palo Alto's (hereinafter referred to by its Chinese abbreviation "Palo Alto Networks") just-raised full-year guidance does not yet include the proposed acquisition of Applied Compute. In the first quarter of fiscal 2027 ended October 31, the company's revenue grew 39% year over year to US$3.445 billion, and adjusted earnings per share rose 13% to US$1.05, both above market expectations of US$3.31 billion and US$0.97. Next-generation security annual recurring revenue (NGS ARR) grew 66% to US$9.73 billion, and remaining performance obligations (RPO) grew 35% to US$21 billion. The company raised its full-year revenue guidance to US$14.52 billion to US$14.57 billion from US$14.1 billion to US$14.2 billion, and raised its adjusted earnings per share guidance to US$4.26 to US$4.30 from US$4.16 to US$4.19. On the conference call, management disclosed two data points that better illustrate the progress of AI monetization than group revenue growth. First, Prisma AIRS, which is specifically designed to protect models and agents, has been commercially available for only five quarters, yet already has more than 1,000 customers and annual recurring revenue approaching US$170 million. Second, agent traffic carried by Palo Alto Networks' secure access service edge (SASE) platform has reached about 9 times the level of the same period a year earlier. The former has already generated subscription revenue; most of the latter is still within capacity customers have already purchased, and inspecting more traffic in real time also increases costs. Palo Alto Networks CEO Nikesh Arora expects that as more customers exhaust their package capacity, excess traffic may begin to generate incremental revenue in the next six to nine months. For investors, this distinction is directly relevant to revenue forecasts: agent usage has already jumped, and revenue growth depends on when contractually agreed capacity thresholds are crossed. The US$4.5 billion all-stock acquisition of Applied Compute is a bet on the next stepthat enterprises will pay to securely run models themselves. This startup provides post-training and inference technology for open-weight models. If the deal closes as planned in the third quarter of fiscal 2027, management expects it to contribute about US$1 billion in next-generation security ARR by fiscal 2029. To support this business, Palo Alto Networks plans to first procure cloud computing power, then add dedicated capacity in batches as demand grows, mainly investing in GPUs and servers. Management's related cumulative capital expenditure figure is less than US$3 billion through fiscal 2030, and it expects the deal to reduce operating margin by 1 percentage point and 2 percentage points in fiscal 2027 and 2028, respectively, at most. Therefore, whether the acquisition can create value will depend on whether the new recurring revenue can cover computing investment and allow margins to recover as scheduled. Whether closed-source or open-weight, the moment agents use enterprise permissions, cybersecurity demand begins to grow substantially A recent series of AI-driven cybersecurity-related attacks shows that attackers can use agents to expand the scope of attacks, and agents authorized by enterprises themselves may also exceed their authority. Palo Alto Networks' Unit 42 disclosed that from August to September 2026, attackers used AI frameworks to conduct 25 operations against at least 19 organizations, of which 7 targets were in the semiconductor and advanced manufacturing supply chain, and 10 organizations were confirmed to have suffered unauthorized access or data acquisition. Another type of risk occurred in October: identity technology company WorkOS confirmed that an agent used by a researcher for security evaluation accidentally escaped its isolated environment and gained unauthorized access to its internal sales tools, affecting data in sales and customer service systems. Both incidents point to the same procurement need: enterprises must be able to identify the caller when anomalous actions occur, limit the scope of its access, and revoke permissions in a timely manner. Regardless of whether the market is ultimately dominated by closed-source models or open-weight models, enterprises must decide what data agents can see and what operations they can perform on whose behalf. For example, an enterprise can authorize an agent to read sales summaries in a cloud data warehouse, but still needs to restrict it from exporting customer-level details or modifying bills. When calling closed-source model APIs, enterprises control these authorizations; when deploying open-weight models themselves, the scope of protection also extends to model files, runtime containers, and inference services. Security vendors that can unify identity permissions, data access, and runtime monitoring therefore have the opportunity to gain more subscriptions as enterprises expand AI deployments. Gartner expects the global market specifically for protecting AI systems to grow from about US$2.835 billion in 2026 to US$4.783 billion in 2027, an increase of 68.7%. For investors, a more direct indicator than betting on which type of model will win is whether security vendors can convert protected identities, applications, and call volumes into sustained payments.