Palo Alto Networks Reports $3.41 Billion in Fiscal Q4 Revenue, Above Analyst Estimates, and Says It Has Acquired AI Agent Startup Console

A Quarter With Two True Numbers

Palo Alto Networks closed the books on its fiscal fourth quarter on September 1, 2026, and handed investors two different bottom lines. Revenue came in at $3.41 billion, up 34% from $2.54 billion a year earlier, beating the roughly $3.35 billion Wall Street expected^[1][2]. Adjusted earnings were $1.02 a share, ahead of the $0.98 analysts had forecast^[2].

But under standard accounting rules, the same quarter produced a $282 million loss, or $0.35 a share^[3]. A year earlier the company had earned $254 million, or $0.36 a share, in that same quarter^[3]. Both numbers are real. They just answer different questions, and neither side of this story disputes either one.

On the same day, Palo Alto said it had bought Console, a startup that builds "agentic" AI, software that takes a goal written in plain English and carries out the steps on its own^[4][5]. The company did not say what it paid^[5]. Console's technology will be folded into Cortex, Palo Alto's security operations product line^[4].

Why a Company Can Grow and Lose Money in the Same Breath

The gap between $1.02 in adjusted profit and a $282 million GAAP loss is not a trick. It's a choice about what to count. GAAP, or generally accepted accounting principles, is the standard rulebook every public company's audited results have to follow. The adjusted number strips out costs the company treats as non-recurring, mostly stock paid to employees and charges tied to buying other companies^[3].

Palo Alto has bought several companies recently, including Protect AI, Chronosphere and CyberArk, and now Console^[6][8]. Each purchase brings expenses that hit the GAAP number but get excluded from the adjusted one. So the loss and the acquisition strategy are, in a real sense, the same decision viewed from two angles.

Management's preferred measure is something called Next-Generation Security annual recurring revenue, or ARR, which reached $9.10 billion, up 63%^[1]. ARR is the yearly value of subscription contracts the company already has signed. It's a real signal of demand, but it's also a number Palo Alto defines itself. No outside auditor signs off on it, unlike the audited net loss sitting in the same release^[1][3].

The company also reported $21.2 billion in remaining performance obligations, contracts already signed whose revenue hasn't been booked yet^[2]. And it guided next year's revenue to $14.10 billion to $14.20 billion, above what analysts expected^[2]. Every one of those guided numbers beat consensus. The GAAP loss just isn't one of them.

The Bill Nobody Disputes

Where the bulls and skeptics genuinely part ways is on cost, not revenue. Gross margin, the share of each sales dollar left after the direct cost of delivering the service, slipped to 74.8%, about a percentage point lower than a year earlier^[8]. Management said cloud-hosting costs will grow faster than revenue in fiscal 2027^[8][11].

That matters because agentic AI products run continuously on rented cloud computing power. Unlike a firewall you sell once and ship, an AI agent that keeps working for a customer keeps costing money to run. The more the product actually works, the bigger the hosting bill gets.

Bullish investors read the AI spending as the price of building what customers are already paying for. Palo Alto reports roughly 2,500 "platformizations," customers who've consolidated onto one of its bundles, with net retention above 120% among that group^[8]. Net retention above 100% means existing customers are spending more each year than they did the year before, even after accounting for anyone who cancels.

Margin-focused investors read the same spending as a warning sign. They point to the audited loss, the rising hosting costs, and a pattern of undisclosed-price acquisitions that keep adding stock-based charges to the books^[8][11]. To this camp, an unspecified price for Console likely means the deal was small, or that the company would rather not say.

A Drop That Happened Before the News Came Out

One number got repeated across headlines that deserves a closer look. Palo Alto shares fell 5.25% during the regular trading session on September 1, then slid roughly another 1.78% after hours^[8][9]. Add them together and it looks like a 7% rejection of the quarter.

It isn't. The earnings came out after the market closed for the day^[2]. The 5.25% daytime drop happened before anyone outside the company had seen the results. Only the smaller after-hours move of about 1.78% was an actual reaction to what got reported^[2][9].

Treating the combined figure as the market's verdict on the quarter mixes up cause and timing. Some outlets, like Investing.com, paired "record growth" with "margin concerns" in the same headline without separating the two moves^[8]. Others, like Seeking Alpha, described shares as "edging up" the same day, a reminder that how a stock move gets framed often depends on which window you're measuring^[10].

Reading the Company's Own Scoreboard

Palo Alto's press release led with growth: 34% revenue growth, 63% ARR growth, framed around the idea that the company is becoming the AI-era consolidator of a scattered security industry^[1]. The $282 million loss appeared further down, in the financial tables^[1][3]. Both figures were disclosed. The order they appeared in is the framing.

Business wires like CNBC and Benzinga largely carried that structure forward, crediting "AI demand" for the beat and giving little space to the loss or the share decline^[2][7]. Security trade press, like SecurityWeek and Network World, mostly skipped the stock story altogether and covered Console as a product, though their descriptions of what it can do came from Palo Alto's own release rather than independent testing^[5][6].

Investor-focused outlets flipped the emphasis toward margins and the share price, treating a same-day stock move as though it settled the question of whether the quarter was good^[8][10][11]. Searches for coverage from Indian, Chinese or Gulf outlets turned up no distinct original reporting on the announcement, only pickups of the U.S. wire and company material. For now, this story is being told almost entirely by Palo Alto itself and by American business media.

What the Number Doesn't Settle

CEO Nikesh Arora bought $10 million of Palo Alto stock on March 27, 2026, his first open-market purchase since 2019, a fact bulls cite as a sign of insider confidence^[13]. In the roughly six months around that purchase, other insiders, led by executive vice president Lee Klarich, sold more than $98 million in shares^[13]. Both facts are on the record. Which one tells you more about where the stock goes next is exactly the kind of question this quarter leaves open.