Palo Alto Networks (NASDAQ:PANW) shares dropped in U.S. premarket trading after the cybersecurity firm posted quarterly results that topped expectations but issued a softer-than-forecast earnings outlook for the year ahead.
Based in Santa Clara, California, Palo Alto provides AI-powered cybersecurity platforms designed to safeguard networks and cloud environments. Its offerings include firewalls, threat intelligence, zero-trust network security, and secure access service edge (SASE) solutions.
According to the company, nine of the Fortune 10 firms are among its clients, along with eight of the 10 largest U.S. banks and six of the world’s 10 biggest oil and gas companies.
In its fiscal second quarter, Palo Alto reported earnings of $1.03 per share on revenue of $2.59 billion, exceeding analyst estimates of $0.94 per share on revenue of $2.58 billion.
“We saw continued strength in platformizations, a trend that is accelerating due to AI – customers are keen to both modernize and normalize their cybersecurity stack, aligning them to our approach. We also saw steady and strong adoption of AI security, which we expect will be a long term trend,” Palo Alto CEO Nikesh Arora said in a statement.
However, the company revised its fiscal 2026 full-year earnings guidance to a range of $3.65 to $3.70 per share, down from a prior outlook of $3.80 to $3.90. The consensus estimate stood at $3.87.
Full-year revenue is now projected between $11.28 billion and $11.31 billion, compared with an earlier forecast of $10.50 billion to $10.54 billion, and above market expectations.
For the current quarter, Palo Alto expects earnings of $0.78 to $0.80 per share on revenue of $2.941 billion to $2.945 billion. Analysts had been forecasting earnings of $0.92 per share on revenue of $2.61 billion.
“[W]e expect AI revenues to inflect meaningfully in revenues over the next two years,” analysts at BofA Securities said in a statement. “Even in a more extreme scenario where AI materially reshapes the security landscape, meaningful disruption remains unlikely in the near term as [large language models] lack the precision needed to displace core security controls.”