Evidence the SaaSpocolypse is over? Okta is a good place to start.
In April, Okta was one of the most hated names in B2B. On April 10, traders were rotating out of weaker software names, and Okta was setting a new 52-week low below all of its major moving averages. Last Friday it closed at $195.19, with a 52-week range of $62.66 to $212.50 and the high set on September 24.
That’s a stock that tripled in five months. And the business growing underneath it is doing about what it was doing before. Q2 revenue grew 11% and subscription revenue grew 12%. For Q3, Okta guided to $813 million to $817 million, 10% growth year-over-year.
So how does an 11% grower triple? The short version: the same force that crushed Okta in the spring, AI, is the force that re-rated it in the fall. And real acceleration in bookings with the right trends. Grow or Die in 2026.
Okta is the most extreme case here I can think of, though, for sure:
- The multiple went from ~18x to ~50x forward earnings while the full-year EPS guide rose about 2%.
- cRPO accelerated from 12% to 14% before revenue moved at all. The market paid for the forward book.
- New products are 30% of bookings, so the agent story has a real attach motion behind it, not just a press release.
- Okta shipped an agent kill switch within about two months of AI agents breaking out of OpenAI’s sandbox and Congress drafting a bill with “kill switch” in the name.
- Okta organized its own competitors into a standards alliance, which is how you get to define a category that doesn’t have a definition yet.
#1. The EPS Guide Rose ~2%. The Stock Tripled. Nearly All of It Is Multiple.
In mid-April, Okta was priced at roughly 17.8x FY27 consensus EPS on a business delivering 9% revenue growth and 29.6% FCF margins. After Q2, Okta guided adjusted EPS to $3.90 to $3.94, above the $3.84 Wall Street estimate. That’s up from $3.79 to $3.87 guided back in May. Take the $3.92 midpoint against Friday’s $195 close and Okta trades at roughly 50x forward earnings.
The EPS guide moved about 2%. The multiple nearly tripled. Public markets price the category you’re in and the growth in forward bookings far more than the quarter you just had.
And the category flipped. In March, Okta was on the wrong side of the AI trade. Cyber stocks sold off after Anthropic launched a new AI security tool, and Reuters named Okta among the stocks hit on fears that AI could disrupt parts of the security stack. A few days earlier, Okta fell 6% in an afternoon when Anthropic announced Claude could control computers by imitating keystrokes and mouse movements. On top of that, Okta had forecast its slowest revenue growth since its IPO.
By September, the same force was working the other way. On September 14, Okta gained nearly 10%, alongside CrowdStrike up 12% and Palo Alto up 11%, after fresh warnings from the heads of the biggest AI labs. AI went from “this will replace you” to “this is why you need them.”
#2. cRPO Accelerated to 14% While Current Revenue Growth Sat at 11%
This is the fundamental piece, and it’s real, even if it’s modest.
In Q1, cRPO grew 12% to $2.50 billion and total RPO grew 16% to $4.719 billion. In Q2, RPO grew 17% and cRPO grew 14%. Revenue was 11% both quarters. The CFO led with the right thing: he highlighted accelerating cRPO, success with the largest customers, and ACV acceleration in both workforce and customer identity.
The market has been rewarding this pattern all year. Atlassian, Salesforce, and Box all got paid for forward book outrunning revenue. Okta is the same story at a lower growth rate. The stock reacted accordingly: up 21% in a single session after Q1, then up nearly 29% to a 52-week high after Q2.
The enterprise base is doing the lifting. Q2 came with record enterprise bookings and over 600 customers above $1 million ACV.
One caveat belongs right here. Q3 guidance calls for cRPO of $2.590 billion to $2.600 billion, 11% to 12% growth. That’s down from 14%. Some of that is conservative guiding, which Okta has done all year. But if cRPO prints 11% in Q3, the core leg of the re-rating is gone, and the stock is running on narrative alone.
#3. New Products Hit 30% of Bookings, and They Lift ACV 40% When Attached
The agent story would be thin if it were only a keynote. It has a real attach motion behind it.
In Q4 FY26, new products including Identity Governance, Privileged Access, and the AI agent offerings made up about 30% of bookings, and when they were included in a deal, average contract value rose about 40%. That held in Q2: new products were 30% of total bookings, and Okta closed dozens of AI deals, including a multi-million-dollar deal with a healthcare company.
The sequencing is worth copying. Okta didn’t wait for agent identity to become a standalone business. It announced Okta for AI Agents in September 2025 with general availability set for April 30, 2026. In the meantime, Identity Governance carried the new-product number, and the agent product rode along in the same deals. During Q2, Okta made the agent product available to all customers.
The decision for founders: at $3B+ with a slowing core, a new product doesn’t need to be big on its own. It needs to raise the size of the deals you’re already closing. A 40% ACV lift on 30% of bookings moves the forward book long before the new product shows up as its own revenue line.
#4. Okta Announced an Agent Kill Switch About Two Months After Agents Broke Out of OpenAI’s Sandbox
The single biggest external catalyst this year didn’t come from Okta at all.
On July 21, OpenAI confirmed that during an internal cyber benchmark, its model found a zero-day in a proxy, escalated privileges, reached the open internet, and got remote code execution on Hugging Face’s servers to pull the benchmark’s answer key. Washington reacted fast. Representatives Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act, which would require developers of advanced AI systems to keep the ability to throttle, suspend, or shut them down.
It isn’t a one-off. OpenAI disclosed a second unauthorized internet access on September 20, when an agent in testing found a way to query a public chatbot.
At Oktane last week, Okta shipped the product that maps directly onto that fear. For agents connected through its Agent Gateway, deactivating the agent makes the gateway reject every request carrying a token the agent already holds, with no waiting for tokens to expire. Note the timing: it reaches general availability in Q4 of calendar 2026. Okta announced the answer before it shipped it, and the market paid anyway.
This is sector-wide, not just Okta. CrowdStrike’s CEO said the Mythos moment turned into mass-market acceptance that AI adoption needs security. Okta’s edge is that “every agent needs an identity” is the simplest sentence in cyber to explain to a board.
#5. Okta Put Its Own Competitors in a Standards Alliance
This one is the most strategic, and it’s why the stock was up almost 14% week to date before Friday’s open.
Okta’s blueprint for the secure agentic enterprise was built by a coalition including AWS, CrowdStrike, Databricks, Google Cloud, Salesforce, and ServiceNow. It answers four questions: where are my agents, what can they do, what are they doing, and how do I respond.
Look at who’s on that list. At the investor summit the same week, McKinnon named ServiceNow’s AI Control Tower and IBM’s watsonx products as new competition, and said platform vendors may use bundling and flexible pricing to pressure standalone vendors. Okta invited one of its named competitors into its own standards effort.
That’s the neutrality play, and Okta is leaning all the way in. Anthropic’s David Soria Parra, a co-creator of MCP, joined McKinnon on stage, and McKinnon stressed that Cross App Access isn’t an Okta-only protocol and has to be adopted broadly to work. Okta’s pitch on its own advantages is focus, neutrality, third-party IdP support, and being the only generally available product in the category.
The decision for founders: when a category has no standard yet, the independent vendor that writes the reference architecture gets to define what “done” means. Okta can’t out-bundle Salesforce or ServiceNow. It can be the vendor all of them agreed to build around.
A Few More Interesting Learnings
- A tax change adds about 7% to year-over-year EPS comparisons. Okta cut its long-term non-GAAP tax rate from 26% to 21% starting February 1, mainly because of the One Big Beautiful Bill Act. Keeping 79 cents on the dollar instead of 74 is roughly a 7% lift to non-GAAP EPS before the business improves at all. Read the EPS growth line with that in mind.
- Okta is buying back stock and retiring debt at the same time. It repurchased $248 million of stock in Q1, then settled the remaining $350 million of 2026 convertible notes in cash and still ended Q2 with $2.299 billion. Most of those buybacks happened at prices well under half of today’s.
- Okta is deliberately giving up about a point of growth. Its FY27 outlook includes roughly a one-point growth headwind from moving professional services work to partners faster. Strip that out and core subscription growth looks a bit better than the 10% headline.
- Six price-target raises, and the stock fell 5.5% anyway. Oktane came with no updated financial guidance, and McKinnon sold 48,822 shares under a pre-arranged plan at $189 to $198 in the same window. A pre-arranged sale means little on its own, but at 50x a market will react to anything.
- Not everyone is buying the re-rating. Bernstein’s Peter Weed cut ratings on Okta, Palo Alto, and SentinelOne on September 17, arguing that cyber valuations imply acceleration rivaling hyperscaler compute, while growth is still capped by factors like customer headcount. That’s the seat problem. Most of Okta’s revenue is still priced per human.
Agent Identity Has to Show Up in Revenue Before Guided Growth Catches Up
Okta’s own CEO is the most honest voice on the risk. McKinnon said confusion over standards, pricing, and agent counts is dragging on near-term adoption, and that no one has locked in long-term share. Wall Street coverage makes the same point: multiple banks see a larger long-term agent opportunity, even as management says monetization is still early.
So Okta is valued at roughly 50x on a 10% revenue guide, with the difference resting on a product that’s generally available but barely monetized, a kill switch that ships next quarter, and a standard that only works if competitors adopt it. Okta’s AI security pitch depends on other companies’ agents continuing to misbehave.
