The quarter. Elastic reported fiscal second-quarter results after the close on Thursday. Revenue was $496 million, up 17%. Sales-led subscription revenue, the contracts Elastic’s sales team signs, was $415 million, up 19%. Current remaining performance obligations, the contracted revenue due in the next 12 months, were $1.157 billion, up 19%. Total remaining performance obligations were $1.858 billion, up 23%. Non-GAAP earnings were $0.94 a share against a $0.81 consensus, per Investing.com. Customers paying more than $100,000 a year rose to over 1,830.
Then the guide. Elastic expects Q3 revenue of $507.5 million to $508.5 million, 12.9% growth at the midpoint, and $2.000 billion to $2.008 billion for the fiscal year. Investing.com put full-year consensus at $2.01 billion.
The tape. Here is the odd part. ESTC had already fallen about 8% in Thursday’s regular session, from $83.01 to $76.33, before a single number was out, per Yahoo Finance. It dropped to around $70 after hours. By about 8:20 a.m. PT Friday it was near $75, roughly 1% under Thursday’s close. The market sold the stock, then read the report, then mostly shrugged.
The bet nobody is talking about. Ignore the stock for a minute. This week the three big ways to build an AI security operations center came into focus.
Palo Alto Networks is buying a team that trains AI models, betting the security vendor should own the brain. We wrote about that yesterday. Microsoft is pricing its SIEM like a feature inside the suite, betting on distribution. We wrote about that on Wednesday.
Elastic is doing the third thing. It is renting the brain. In the quarter it said it plans to bring OpenAI’s cyber models into Elastic Security, and it joined Anthropic’s Cyber Verification Program. Its new SOC agents, called AlertZero, work with whatever model a customer picks, and analysts can switch models halfway through an investigation. They run in Elastic’s cloud, on a customer’s own servers, or fully air-gapped. Every consequential action, like isolating a laptop, is proposed to a human first, and detection rule changes always wait for approval.
What Elastic keeps is the data and, now, the report card. It agreed to buy HoneyHive, a startup that traces what AI agents do and grades whether they did the right thing. Terms were not disclosed. HoneyHive’s CEO put the logic simply: explaining why an agent failed is increasingly a search problem. Search is what Elastic does.
Why we think Elastic has the better idea. Models are the part that changes every quarter. The best cyber model in January may be third by June. A SOC that bets on one vendor’s model is locked into that vendor’s pace. A SOC that owns clean, searchable data and a way to grade every agent’s work can swap brains when a better one shows up. For a buyer, that is the lower-regret architecture, especially for shops that must run air-gapped, which AlertZero supports.
Elastic also has a proof point on the defense side. It says it got the only perfect protection score in AV-Comparatives’ 2026 endpoint prevention and response test, stopping all 50 attack scenarios at the first stage with zero false positives. That is Elastic citing a third-party test; it is worth reading the report yourself.
Why the market isn’t paying for it yet. The model-neutral story is good for buyers. It is harder for investors, because it is harder to charge a premium for the layer under someone else’s AI. And the numbers don’t show the AI lift yet. Revenue grew 17%. The Q3 guide is 12.9% reported and 13.9% in constant currency. Elastic says a stronger dollar takes $4.5 million out of Q3. That explains about one point of the slowdown. It does not explain the rest.
The gap we would watch: contracted revenue for the next 12 months is growing 19%, total backlog 23%, and next quarter’s revenue guide is under 13%. Either management is guiding very carefully, or customers are signing bigger, longer deals and using them more slowly. AlertZero is still a technical preview. HoneyHive hasn’t closed. Neither is in the numbers.
Bottom line. Palo Alto wants to own the AI. Microsoft wants to give it away. Elastic wants to own what the AI reads and how it gets graded. For the buyer, that is the most flexible bet on the board. For the stock, it needs the backlog to show up as revenue.
