Two security companies reported on the same evening and both stocks had an extraordinary day. CrowdStrike closed up 20.5%. Okta closed up 28.6%.
They sell different things. They gave the same explanation.
The numbers
| CrowdStrike | Okta | |
|---|---|---|
| Revenue | $1.47bn, +26% | $805m, +10.6% |
| Subscription revenue | $1.40bn, +27% | $793m, +12% |
| Adjusted EPS | $0.31 | $1.05 |
| Headline metric | Net new ARR $332.8m, +51% | FY guide raised to $3.22bn |
| Close, 27 Aug | $227.96, +20.5% | $172.91, +28.6% |
The CrowdStrike line that did the work is net new ARR accelerating 51% year on year to a record. Total ARR growth of 25% is good; net new ARR accelerating at a company already at $5.84 billion is the thing that does not usually happen, and it is why the reaction was that size.
Okta's growth is half CrowdStrike's — 10.6% is not a hypergrowth number — and it still rose more. That reaction was about the guidance raise landing on a stock nobody was positioned in.
Why the same driver appears twice
Both managements pointed at AI agents.
An agent that books travel, files tickets or moves money needs credentials, permissions and system access. It acts without a person approving each step. Every agent is therefore an identity, and identities in the tens of thousands are not something an access-review spreadsheet handles.
That is a straightforward budget consequence. Deploying agents at scale creates a security cost that scales with the deployment. It is not discretionary spend that gets cut in a soft quarter; it arrives attached to a project the board has already approved.
Which is why these two prints are, structurally, AI capex — just landing on a different set of income statements than the semiconductor ones.
The wider pattern in the week
Salesforce rose 22.6% the same day for converting AI into margin now. CrowdStrike and Okta rose for selling into a budget that AI adoption creates now. Nvidia rose for a forecast about fiscal 2028 and gave half of it back within a session.
The market spent this week paying for the near end of the AI trade and discounting the far end. With September rate-hike odds moving to a coin flip on Friday, that is exactly the sorting you would expect.
How I read it
The second-order beneficiaries of AI are having a better year than the story suggests, and they are cheaper to be wrong about than the chip makers.
What I would watch on Okta specifically: 10.6% revenue growth does not support a 28.6% single-day move on its own. That was a repricing of expectations, not of the business, and repricings of expectation reverse faster. It already gave back 3.9% the following session.
CrowdStrike's number is the more durable one. Net new ARR is a forward measure, and it accelerated.