Elastic finally throws off cash — just as GenAI makes its engine matter

• The cash inflection: free cash flow margin climbed from 3% to 18.5% in three years — $33M to $322M — and FY2026 brought the company's first-ever buyback.
• The new relevance: rebranded "the Search AI Company," Elasticsearch is now positioned as a leading open-source vector database, the retrieval layer that feeds context to large language models.
• The one question: can Search AI keep revenue growing in the mid-to-high teens and widen margins together — or is ~3x revenue the right price for a steady compounder?

Two gauges point opposite ways — and the stock waits on which one leads

• The trailing gauge is frozen: Net Expansion Rate has held at ~112% for eight straight quarters, down from ~130% in FY2022 — existing customers have largely stopped scaling spend.
• The forward gauge just jumped: current remaining performance obligations accelerated to +20% constant-currency in Q4 FY2026, and total backlog to +27%, while reported revenue grew 16%.
• AI is now real demand, not a label: 600+ customers spending over $100k a year run AI use cases — a third of that cohort, up from 17% a year earlier — and that cohort grows about 5 points faster than the base.

The 18% cash margin thins sharply once stock pay is charged as a cost

Stock-based compensation is 97% of the bridge from Elastic's GAAP operating loss to its non-GAAP profit, and is roughly the same size as reported free cash flow. Charge it back as a real cost and ~$322M of FCF falls toward ~$24M. The trend is the bull's friend — SBC eased to ~17% of revenue and the first buyback finally cut the share count — but about $280M of that buyback merely sterilized dilution rather than returning capital.

At ~$57 the tape prices the bear; re-acceleration is a near-free option

Elastic carries the lowest revenue multiple and highest cash-flow yield in its cohort — but also the slowest growth, so the discount reads as the market's verdict, not its mistake. The downside is shallow: a net-cash balance sheet plus precedent take-outs (Splunk ~7x, New Relic ~6.5x revenue) imply an ~$80-plus private-market floor — though a Dutch foundation poison pill and ~80% insider control make that floor collectible only if the founders choose to sell.

Defensibility lives in data gravity and a vector engine, not the license

• The license was given away twice: Elastic relicensed in 2021 to block cloud resellers, AWS forked OpenSearch within weeks, and Elastic added an open AGPL option back in 2024 — that fork is a permanent overhang.
• Out-spent on every front: a sub-scale generalist fighting Datadog, CrowdStrike and the hyperscalers — Datadog alone spends ~$1.55B on R&D against Elastic's $452M.
• What actually defends it: 5.8B cumulative downloads and sticky data gravity, plus a vector engine management claims runs up to 8x faster than OpenSearch, anchoring 2,700+ cloud vector-database customers.