Product1 distinct publisher3 min readPublished
CrowdStrike and Okta both credit enterprise agent rollouts for larger commitments. The commitments moved sharply; the quarters they reported barely did.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
The two beats are the wrong shape for the price action. CrowdStrike's revenue of $1.47 billion, up 26%, came in about 2% above the $1.44 billion consensus [3][4][2]. Okta's $805 million, up 11%, landed about 1.6% above the $792.1 million analysts wanted [16][17][3]. Those are ordinary beats. What was not ordinary sat in the forward book: CrowdStrike's net new ARR cleared the midpoint of the range it gave in June by 16.8% [1], and at Okta the subscription backlog is outgrowing revenue, with the portion stretching beyond the next twelve months up roughly 21% year over year [4]. Customers are committing to more than they will consume this year, which is what you would expect if they are sizing for agent fleets that are not yet in production.
Neither company quantified that, and CrowdStrike's largest supporting disclosure makes the attribution harder rather than easier. Accounts running on Falcon Flex, the flexible licensing arrangement pushed since 2024, carried more than $2.29 billion of the $5.84 billion ARR total and grew 101% [7][6], about 39% of the book [6]. Back out the year-ago figures implied by those growth rates and Flex accounts hold roughly 98% of the $1.17 billion of ARR added over twelve months [7]. Flex is counted per account, so a customer that shifts existing spend onto the vehicle brings that spend into the bucket with it. The metric records where ARR is booked, not why it grew. theCUBE Research's Dave Vellante read the record net new ARR as AI threat urgency converting into demand [13], then cautioned that customers do not buy contracting vehicles and that Flex is the commercial harness rather than the engine [14].
The more durable evidence is in the product line. Continuous Identity for AI Agents extends risk-aware authorization across human, machine and agent identities [11]. An agent that has to be authorized is an agent that can be counted, and counted things get priced. That is the mechanism by which an agent programme turns into a recurring security and identity line rather than a pilot, and unlike management commentary it will show up in a unit metric someone can check.
One figure worth holding onto: unadjusted net income was $5.3 million on $1.47 billion of revenue [8][3], a margin of 0.36% [8], in the quarter George Kurtz called the best in CrowdStrike's history [10]. Operating cash flow of $530.3 million and free cash flow of $377.4 million were second-quarter records [9], so the cash engine is real. The distance between 31 cents adjusted and a penny unadjusted [3][8] is where the cost of buying this growth is sitting, and it is not disclosed as an agent-security investment either.
Ranked by verification strength, evidence, and original report placement.
Executives at both CrowdStrike and Okta pointed to enterprise deployment of artificial intelligence agents as the force pulling customers toward larger security commitments.
Accounts running on Falcon Flex, the flexible licensing arrangement CrowdStrike has pushed since 2024, carried more than $2.29 billion of total ARR, and their annual recurring revenue grew 101%.
Dave Vellante, chief analyst at theCUBE Research, said record net new ARR of $333 million, up 51% year over year, indicates that the urgency created by AI-powered threats is converting into significant demand.
Vellante also said customers do not buy contracting vehicles, they buy outcomes, and that Falcon Flex "may be the commercial harness, but customer value is the real engine we should pay attention to."
CrowdStrike shares were up more than 10% after the bell and Okta shares more than 19%; both companies raised their full-year outlooks.
For its fiscal 2027 second quarter, ended July 31, CrowdStrike reported adjusted earnings of 31 cents per share, up from 23 cents a year earlier, on revenue of $1.47 billion, up 26% year over year; per-share figures reflect a four-for-one stock split completed in July.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Strong reported financials, thin independent corroboration
Every quantitative claim traces to company earnings releases with precise, checkable figures (revenue, adjusted and unadjusted EPS, ARR, net new ARR versus prior guidance, RPO, cash flow), and the derived arithmetic follows directly from those disclosures. Evidence is capped because the cluster has a single trade publication, the only interpretive voice is that publisher's sister research firm, and the central narrative claim — AI agent deployment as the cause of larger commitments — rests entirely on management assertion with no deployment or consumption data.
Commitments measurable, agent usage not
Adoption is well documented at the level of contracts and backlog: $332.8 million net new ARR, $5.84 billion total ARR, more than $2.29 billion of ARR on Falcon Flex growing 101%, and Okta RPO of $4.858 billion with the current portion at $2.585 billion. What is absent is any measure of the thing the story is about — agents actually deployed, identities issued, or capacity consumed. Committed ARR and backlog growing two to five times faster than recognized revenue is consistent with buying ahead of deployment, so adoption of the underlying agent-security workload cannot be scored higher than the contracting evidence supports.
Commitment metrics outrun delivered results
The gap is arithmetic rather than rhetorical. Net new ARR grew 51% and cleared guidance by about 17% while revenue grew 26% and beat consensus by about 2.1%; at Okta revenue grew 11% while longer-dated backlog grew about 21%. CrowdStrike's GAAP net income of $5.3 million is a 0.36% margin. Against that, the CEO calls it 'the best quarter in CrowdStrike's history' and 'the largest market opportunity in our history,' and the AI-agent causal story is asserted, not evidenced. Cash flow records and the concrete raised guidance keep the overstatement moderate rather than severe.
Promotional sourcing with a disclosed affiliate analyst
Nearly all narrative material originates with parties that benefit from it: CrowdStrike's and Okta's own earnings releases and executive quotes, published into an after-hours move of 10% and 19% respectively, with CrowdStrike stock already up close to 70% for the year. The AI-agent explanation is supplied by the sellers of agent-security products. The only interpretation is from theCUBE Research, which the publisher discloses is its own sister research firm — a disclosed but real alignment. Partly offsetting: that analyst voices a counterweight, warning that Falcon Flex is a contracting vehicle rather than the value engine.
Figures reliable, interpretation single-sourced
Confidence is high on the numbers — they are specific, internally consistent, and drawn from earnings releases whose arithmetic checks out — and moderate overall because the cluster contains one publisher, the analytical read comes from an affiliated firm, and the causal claim at the heart of the story is unverifiable from the supplied material. Judgments about the commitment-versus-delivery gap are robust; judgments about what is driving it are not.
leadership
CrowdStrike buys SGNL, and standing privilege becomes a line item you have to defend1 distinct publisher
build
An agent built and deleted a prod stack. The alert fired on time and changed nothing1 distinct publisher
product
One unvalidated region string sent signed AWS API calls to attacker.com1 distinct publisher
product
Stealth is now a launch strategy: Zhipu's Ox Alpha topped the charts before it had a name1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 26, 2026