Have you ever watched a stock spend months looking broken, only to snap higher the moment the numbers finally match the story the company has been telling? That is pretty much what happened with Zscaler after its fiscal fourth-quarter report. The shares jumped once investors saw a clean beat, firmer guidance, and a chief executive who sounded more confident talking about artificial intelligence than he has in a while. I have covered enough earnings seasons to know a one-day pop is not a thesis. Still, this print felt like a reset after a year in which cyber peers made new highs and this name quietly slid about twenty percent.
Why Zscaler Stock Surged After The Fourth Quarter Print
The market did not need poetry. It needed proof. Adjusted earnings came in at $1.19 a share against an expectation near $1.09. Revenue reached $898 million, ahead of the $877 million consensus and up about 25% from roughly $719 million a year earlier. Annual recurring revenue climbed to $3.77 billion, a shade above the $3.75 billion figure many desks were using. None of those gaps is enormous on its own. Together they told a simpler story: demand is still there, execution is not falling apart, and the guide is not being sandbagged into oblivion.
On a GAAP basis the company still lost money. The net loss narrowed to about $3.37 million, or two cents a share, from a loss of $17.58 million, or eleven cents, a year ago. That is not the headline investors traded. They traded the adjusted run-rate, the billings pulse, and the fact that management did not hide behind another round of “prudent” language after last quarter’s ugly session.
The core competency we bring to the table is pretty unique, and as the adoption of AI agents happens, the market will recognize more and more that Zscaler is a critical player.
– Company leadership on differentiation
I found that line more interesting than the beat itself. Peers have been rewarded for any AI-adjacent sentence. Zscaler had the opposite problem. The stock lagged even while the industry narrative got louder. Management’s view is that the market has been pricing the company like a standard secure-access name instead of a platform that sits in the path of agent traffic. Whether that is fair is another debate. After this quarter, at least the numbers stopped arguing against them.
The Quarter In Plain Numbers
Let us put the scoreboard in one place. Short paragraphs help when the figures start stacking up, and this table is the cleanest way to see why the tape moved.
| Metric | Reported | Street View | Takeaway |
| Adjusted EPS | $1.19 | $1.09 | Clear beat |
| Revenue | $898 million | $877 million | Solid top-line surprise |
| Year-ago revenue | About $719 million | — | Roughly 25% growth |
| ARR | $3.77 billion | $3.75 billion | Slight beat, still 25% up |
| GAAP result | Loss of $3.37 million | — | Narrower than last year |
Growth at twenty-five percent is not the hypergrowth of five years ago. It is also not a stall. In security software, that mid-twenties clip with expanding platform attach can still support a premium if the mix is shifting toward stickier modules. That is the bet buyers made on the print. Sellers had spent months assuming the opposite: that sales leadership turnover and a cautious guide last quarter were early signs of a longer fade.
Guidance That Finally Stopped The Bleeding Narrative
Guidance is where this story gets practical. For the first quarter, the company pointed to revenue of $935 million to $939 million and adjusted earnings of $1.15 to $1.16. The Street had been closer to $927 million and $1.08. For the full year, revenue is expected between $3.91 billion and $3.94 billion versus a $3.90 billion estimate. Adjusted EPS is slated for $4.86 to $4.90 against something nearer $4.60.
That EPS gap is the one that jumped off the page for me. A few cents on the quarter can be mix and buybacks. A full-year lift of that size suggests operating leverage is showing up in the model, not just in a single clean month of billings. Of course, guidance is still guidance. Security buyers can freeze. Deals can slip. A handful of large renewals can distort a quarter. Even so, after last quarter’s shock, beating the guide and then raising the bar was the minimum requirement to get the stock a hearing again.
- Near-term revenue guide sits above consensus, not merely in line.
- Adjusted EPS for the year is marked well ahead of the prior Street number.
- ARR growth remains aligned with the top line, which matters more than a one-off product spike.
- GAAP losses are shrinking, even if profitability is still an adjusted conversation.
In my experience, the market forgives a lot when the next twelve months look better than the model. It punishes companies that keep using the word prudent after they have already taken the medicine. This time the tone was closer to “we see the pipeline” than “please do not get ahead of us.” That change in register is hard to quantify and easy to feel on a trading desk.
Zero Trust Was The Engine, AI Agents Are The New Hook
Management credited the Zero Trust cloud architecture and a run of product work for the beat. That is the familiar pitch. The fresher pitch is agentic security. The chief executive said he is very bullish on a recently launched platform for AI agent iteration and described it as a potentially larger long-term annual recurring revenue opportunity. Early momentum is there, he argued, with acceleration expected into 2028 and 2029.
It is a longer-term opportunity, but I think it is a fantastic opportunity with significant barriers to entry.
That last phrase is doing a lot of work. Barriers to entry in security are never absolute. What companies usually mean is that you need traffic visibility, identity context, policy engines, and a data lake that already watches how users and workloads behave. If you already sit on that stream, wrapping controls around autonomous agents is an extension. If you do not, you are bolting a new product onto someone else’s map of the enterprise. I tend to agree that incumbents with a real inspection point have an edge. I also think investors should treat 2028 commentary as a horizon, not a coupon they can cash next spring.
Over the last year, bookings tied to this AI security push totaled about $100 million and have been growing roughly 50% sequentially, quarter over quarter, according to the company’s own framing. That is still small against a multi-billion ARR base. It is not small as a proof point. Sequential fifty percent on a new motion is the kind of curve product teams dream about, provided the definition of those bookings stays honest and does not quietly absorb ordinary secure web gateway deals wearing an AI sticker.
The Industry Backdrop: Agents, Attacks, And A Crowded Trade
Cybersecurity names have had a roaring year. The simple explanation is demand. Companies are standing up swarms of AI agents, and every agent is another identity that can be hijacked, prompt-injected, or used as a lateral-movement tool. The less simple explanation is that attackers are using the same class of models. You do not need a thriller plot to see why boards suddenly care. A compromised agent with production credentials is not a phishing email. It is an automated insider.
That backdrop helped peers print new highs while Zscaler went the other way. Last quarter the stock had one of its worst sessions after leadership changes in sales and a cautious stance on the outlook. Markets hate two things at once: a people story and a guide that sounds like a shrug. This quarter tried to close both files. The numbers improved. The AI commentary got specific enough to sound like a product, not a slide.
Is the sector frothy? Perhaps. I have found that security spending is one of the last lines finance teams cut and one of the first they regret cutting. That does not make every multiple sane. It does mean the end market is real. When executives across the group talk about AI security as one of the biggest openings in a decade, they are not inventing a category from nothing. They are pointing at a change in the attack surface that even skeptical operators can see in their own labs.
What The Market Got Wrong, At Least In Management’s Telling
The stock was down about twenty percent on the year heading into the report, even as the broader cyber complex ran. Management’s answer is differentiation. The claim is that Zero Trust plus a cloud-native inspection fabric is not interchangeable with a pile of point tools. As agents proliferate, the argument goes, buyers will care who can enforce policy on machine-to-machine traffic without hairpinning everything back through a castle-and-moat design.
I will be honest. Every vendor says they are not a point tool. The test is attach rates, net expansion, and whether large customers consolidate onto the platform after the first module. If those metrics hold through the next two prints, the “misunderstood” narrative gains weight. If they wobble, the lag versus peers will look like a verdict rather than a misunderstanding.
- Watch net expansion, not just logo growth.
- Separate AI-tagged bookings from the core secure access motion.
- Track sales capacity after the leadership turnover.
- Compare win rates in large platform deals versus single-product swaps.
- Listen for any change in the duration of the sales cycle.
Those five items are more useful than arguing about a one-day percentage move. A surge after a beat can fade by the following Friday. The operating questions stay.
How To Read Recurring Revenue When Growth Cools From The Old Peak
ARR of $3.77 billion, up twenty-five percent, is the heartbeat. In subscription software, ARR is the thing you can almost take to a bank, provided churn is contained and discounting is not doing secret damage. A two-cent GAAP loss next to that ARR base is a reminder that reported profit and economic engine are still different conversations at this stage of the model.
Why does that matter for valuation? Because many holders are not buying last quarter’s earnings. They are buying the slope of the remaining performance obligation and the chance that AI modules lift average revenue per customer. If ARR growth were sliding into the teens while guidance stayed timid, the multiple would compress fast. It did not slide. That is the quiet win underneath the louder AI quotes.
A simple way to frame the print: Growth still mid-twenties Guide above the model AI bookings small but accelerating GAAP loss shrinking Stock still working off a lost year of relative performance
Notice what is missing from that sketch: a claim that the shares are cheap on every measure, or that the competitive field has suddenly emptied. Neither is true. The field is crowded. Palo-style platform wars, identity specialists, and incoming AI-native startups all want the same budget line. Zscaler’s case is architectural. You either buy that case or you wait for proof in mix.
Investor Psychology After A Lost Stretch Versus Peers
There is a human layer here that models ignore. When a group rips and one name does not, holders start telling themselves a punishment story. Maybe the product is late. Maybe the sales force is broken. Maybe the multiple was a relic. Those stories can be right. They can also overshoot. Last quarter fed the punishment story. This quarter poked holes in it.
I have found that catch-up rallies in software often last longer than skeptics expect when two conditions line up: the print removes a specific fear, and the sector bid is still alive. Both were present. The specific fear was that guidance would again disappoint after sales-leader noise. The sector bid is the AI-security trade. That combination can squeeze people who used the name as a pair trade against stronger peers.
Does that mean chase the gap on the open? Not automatically. Liquidity after earnings can be theatrical. The better habit is to ask whether your thesis needed this beat. If you already believed the platform story, the report is confirmation. If you needed the report to invent a thesis, you are late to the emotion and early to the hard work of modeling 2027.
The Agentic Opportunity Without The Slideware
Let us talk about agents in language that would survive a security review, not a keynote. An agent is software that can plan and act. It calls tools. It holds context. It often holds secrets. Traditional controls assumed a human at a keyboard and a browser in the middle. Agents skip the browser. They speak APIs. They spawn child tasks. They persist.
That is why a Zero Trust fabric that already brokers connections between identities and resources has a natural on-ramp. You can treat an agent as another identity, issue it least privilege, inspect the calls, and kill the session when behavior drifts. Easy to say. Hard to do if you lack the broker. That is the “barriers” point again, minus the marketing frosting.
The timeline into 2028 and 2029 is actually refreshing. Too many firms imply the AI dollar lands this winter. Large enterprises do not rip identity and network patterns in a quarter because a demo looked pretty. They run pilots, argue with legal, and then expand. A multi-year ramp is more believable than a hockey stick drawn on a napkin.
Security for AI is one of the biggest opportunities in the category, but the dollars follow architecture, not adjectives.
Perhaps the most interesting aspect is sequencing. First you secure the humans using AI tools. Then you secure the tools acting as humans. Then you secure the agents that hire other agents. Each layer is a new policy problem. Vendors who already store the telemetry will train faster. Vendors who only sell a scanner will live on the edge of the budget.
Risks That Did Not Vanish Because The Quarter Was Clean
A good print does not retire risk. Competition remains intense. Customers can dual-source. Discounting can sneak into large renewals. A broader IT spending pause would hit even “must have” categories, just later than it hits discretionary software. There is also execution risk around the new agent platform. Early momentum can stall if the first cohort of design partners hits friction in production.
Sales coverage is another live wire. Two leader departures were enough to spook the market once. The organization has to show that productivity per rep is stable. You cannot see that in a single quarter of revenue. You see it in pipeline coverage comments and in the shape of large-deal contribution over a few periods.
- Competitive bake-offs can compress price even when volumes hold.
- AI bookings need a durable definition so the 50% sequential claim stays apples to apples.
- Macro shocks still move security refresh cycles, just with a lag.
- Valuation can re-rate down if the whole cyber group cools, regardless of this print.
I would rather hold those risks in the open than pretend a beat erased them. Adults can own a stock and still list the ways the story breaks. That is not bearish. That is hygiene.
What A Reasonable Forward Case Looks Like
A base case is not heroics. It is mid-twenties revenue growth cooling only gradually, ARR staying in step, and the AI module becoming a visible but not dominant slice of new bookings by late next year. In that world, the raised EPS range starts to look less like a stretch and more like the operating model catching up to scale.
A bull case needs two extra things. First, platform attach has to accelerate so customers buy more than the original on-ramp product. Second, the agent story has to show up in win/loss notes against rivals who are also waving AI flags. If both happen, the multiple can heal the underperformance versus the group without requiring a fantasy about unlimited total addressable market.
A bear case is simpler and uglier. Growth steps down faster than the guide implies, AI bookings flatten after the anniversary of the launch, and sales execution noise returns. In that version, this rally is a short-covering event, not a regime change. You do not need to pick a camp tonight. You need to know which data would force you to switch camps.
How This Fits A Broader Portfolio Conversation
Not every reader is trading the name. Some are asking whether cloud security still belongs in a growth sleeve after a noisy year. My view is mildly old-fashioned. Security remains one of the cleaner ways to express digitization without betting on a single consumer app. The caveat is concentration. Owning five names that all rise and fall on the same board-level cyber budget is not diversification. It is a theme with extra tickers.
Position sizing after a gap-up is where people get sloppy. Adding because the chart looks like relief is not a process. Adding because the next twelve months of revenue and earnings now sit above your model is a process. If you did not have a model, the article you want is not a victory lap. It is a homework list: unit economics, remaining performance obligations, large-customer concentration, and the competitive matrix in Zero Trust plus data protection.
There is also the question of time horizon. The CEO’s own words pushed the fattest part of the agent opportunity into 2028 and 2029. That is a long time in equity land. Quarterly traders will not care. Compounders might. The mismatch between those clocks is how good companies stay mispriced, and also how patient holders get bored and leave before the second act.
A Closer Look At The Tone From The Top
Tone is not a number, but after last quarter it became a number in disguise. The language this time was assertive. Unique core competency. Critical player as agents spread. Fantastic opportunity. Those phrases can sound like filler until you stack them against the prior posture. The earlier message felt like a company asking the market for time. This one felt like a company asking the market for a second look.
I do not treat executive confidence as evidence. I treat it as a tell about what they think the debate is. Right now they think the debate is differentiation, not survival. That is a better debate to be in. It is also a debate they can still lose on the merits if rivals match the agent story with similar bookings growth.
One subtle opinion: founders who still run the company often sound most alive when the product cycle turns. You could hear that here. The risk with founder energy is over-identification with a single architecture. The benefit is that they will not quietly harvest the installed base and call it a strategy. For a platform in mid-innings, I will take the second problem over the first.
Putting The Beat In Historical Context Without Nostalgia
There was a period when twenty-five percent growth would have disappointed this shareholder base. That period is gone. Rates are not the same. Public multiples are not the same. The easy lift from “everyone is leaving the castle” is mostly behind the category. What remains is the harder work of replacing a patchwork of tools and then selling the next control plane for machines that act on their own.
Seen that way, this quarter is less a return to the old religion and more a demonstration that the new religion still pays the bills. Revenue up a quarter on a year. ARR up a quarter on a year. Losses smaller. Guide higher. AI bookings no longer theoretical. That is a grown-up print. Grown-up prints do not always produce grown-up valuations the next morning, but they do give serious investors something to underwrite.
Practical Takeaways If You Follow The Name
If you only remember a handful of points, make them operational. The beat was real across earnings, revenue, and ARR. The first-quarter and full-year outlook cleared the bar that last quarter missed in spirit if not always in math. The strategic conversation has shifted toward securing AI agents, with a booking figure that is still modest and a timeline that is honestly long. The stock’s year-to-date lag versus peers created dry powder for a squeeze once those facts landed.
What you should not remember is a cartoon. This is not a company that suddenly has no competitors. It is not a GAAP profit machine overnight. It is not a guaranteed winner of the agent era. It is a scaled cloud security vendor that just reminded the market it can still outrun a cautious model while talking credibly about the next platform layer.
- Mark the guide ranges and revisit them at the next print rather than treating them as destiny.
- Keep an eye on sequential AI bookings without letting them overshadow core ARR.
- Respect the sector tailwind and the single-stock lag as two different facts.
- Refuse to confuse a relief rally with a completed rerating.
- Write down, today, what would make you wrong.
That last habit sounds small. It is not. After a sharp up day, brains rewrite history. People decide they “knew” the beat was coming. A written checklist survives that rewrite. Mine would include sales productivity, large-deal timing, and whether the agent platform remains a distinct motion or becomes a slogan painted on every contract.
The Longer Arc From Cloud Access To Machine Identity
Step back far enough and the product history is almost linear. First you secure users leaving the office. Then you secure apps leaving the data center. Then you secure workloads talking to other workloads. Now you secure software that can choose its next action. Each step added identity, context, and inspection. Each step made the old perimeter look more like a souvenir.
Zscaler’s wager is that it already lives at the intersection of those steps. If that wager is right, agent security is not a side quest. It is the same inspection point with new principals. If the wager is wrong, agents will pull security spend toward vendors who start from model governance and work backward into the network. I lean toward the first outcome for large enterprises that already standardized on a broker. I would not bet the house against the second for greenfield AI stacks inside startups that never built a traditional access layer.
That split is where the 2028 and 2029 comments become useful. They imply a world in which agents are common enough inside Fortune-scale estates to generate real ARR, not just design-partner logos. We are not there yet. We can see the hallway from here. Walking it will take budget cycles, incident-driven urgency, and a few public failures that make boards stop treating agent controls as optional.
Final Thoughts After The Tape Settles
So where does that leave a reader who is not paid to stare at Level 2? It leaves you with a cleaner fundamental snapshot than the one you had yesterday. The company beat. It guided higher. It put a number on AI-related bookings and a calendar on the bigger prize. It also reminded everyone that last year’s relative slump was not the same thing as an operating collapse.
I keep coming back to a simple contrast. The industry spent a year celebrating anything that touched artificial intelligence. This name spent a year being treated like last cycle’s workhorse. Workhorses can look dull until the road turns rough and you need something that already knows the route. Agent traffic is a rough road. Whether Zscaler is the right horse is still a living question. After this quarter, it is at least a fair question again.
If the next report shows the same pattern — growth holding, guide intact, agent bookings still compounding — the misunderstood-story stops being a speech and starts being a tape. If it does not, this surge will join the pile of earnings pops that felt profound on Thursday and ordinary by the following month. That tension is what makes the name watchable. It is also why a single strong quarter should be filed as evidence, not as a conclusion.
For now, the facts are stubborn in a helpful way. Revenue of $898 million. Adjusted earnings of $1.19. ARR of $3.77 billion. A first-quarter revenue band of $935 million to $939 million. A full-year earnings band of $4.86 to $4.90. A hundred million in AI-linked bookings over the past year, rising fast from a small base. Those are the bones. The rest is interpretation, temperament, and time. I would rather argue about those bones than about a chart that spent most of the year going the wrong way for reasons that finally have a chance to be outdated.
