SAP reported Q1 2026 cloud revenue of €5.96 billion, up 27% at constant currencies, with a current cloud backlog of €21.9 billion — also up 25% CC. Non-IFRS operating profit rose 24% CC. Five weeks earlier, CEO Christian Klein had called subscription pricing “foolish”, because AI would automate the tasks seat licenses charge for. Both statements are true, and that is the interesting part.
The Q1 print is the strongest case for hypothesis falsification since this blog opened — a legacy ERP vendor growing cloud revenue at a pace well above the market while pivoting, out loud, to consumption pricing. Cloud gross margin held flat at 74.6%. Business AI was reportedly in two-thirds of Q4 cloud orders and in 90% of the fifty largest deals. Daimler Truck North America cited 4x higher bid win rates. Bosch Digital equipped 1,500 developers with SAP’s AI tooling for the ERP migration. This does not look like erosion.
Look closer at where the revenue came from.
SAP maintains roughly 10,000 on-premise ECC customers facing a 2027 end-of-maintenance deadline. Joule — SAP’s AI layer — only runs on RISE or GROW cloud contracts. That means the migration cliff and the AI story are the same story: customers who were always going to re-paper into cloud contracts by 2027 are doing so now, with AI components attached because they come with the package. Software licenses revenue — the perpetual base being drained into that cliff — fell 37% to €116 million. Software support fell 11%. Management now expects the support-revenue decline to accelerate, not stabilise.
Then there is the gap between attach and use. SAP’s own narrative says Business AI is in two-thirds of Q4 cloud orders. The DSAG 2026 Investment Report — surveying the German-speaking SAP user group — finds that only 3% of SAP customers run SAP Business AI productively. Seventy-seven percent of SAP shops using AI at all are using Microsoft Copilot or other non-SAP tools. Those two figures are reconcilable if AI is getting checked into contracts during the cloud move without actually running in production. Attach rate is a commercial indicator; production use is a value indicator. They diverge here by more than an order of magnitude.
The guidance language shifted too. SAP still points to €25.8–26.2 billion in cloud revenue for 2026. But total revenue growth is now expected to “remain at similar levels as in 2025” rather than to accelerate through 2027 — a downgrade embedded inside a maintain. Cloud revenue growth is expected to decelerate in Q2. And management acknowledged that the March acquisition of master-data-management firm Reltio was rolled into guidance to protect the range against an underlying services setback this year. Q1 total revenue came in slightly below the sell-side consensus even as cloud beat.
On the AI Units model itself: Joule is charged at €7 per AI Unit, with a 100-unit minimum, overages at 150–200% of the contracted rate, and volume-tiered discounts. That is a consumption curve that looks nothing like per-user. It will take several quarters to see whether the unit economics — at 3% production adoption — deliver anything close to the revenue density of the seat licences it replaces. SAP is, in effect, running two monetisation engines at once: per-user contracts still recognised in revenue, consumption-priced AI stapled on top, with the CEO publicly betting on the second and the income statement still carried by the first.
This complicates the hypothesis more than it verifies or falsifies it. The “software loses value because AI can create it fast and efficiently” argument has always had a specific weak point around incumbent workflow-gravity platforms — SAP is the textbook example. SAP’s Q1 beats the strong version of the hypothesis: the numbers are not collapsing, margins held, backlog grew. But it verifies a narrower version: the vendor with the deepest moat now publicly treats its own pricing model as obsolete, reports accelerating erosion of its legacy licence and support base, and shows a 25x gap between AI attach on paper and AI in production. The growth is real and probably not repeatable once the 2027 migration wave runs out.
What to watch: SAP Sapphire on May 13 for AI Units volume curves and any attach-to-production metrics management is willing to disclose. Q2 earnings in July for the promised cloud revenue deceleration. Any subsequent DSAG reading that moves the 3% production-use figure. And whether Q2 software-support revenue declines by more than the 11% Q1 rate — the speed of that bleed, more than the cloud growth rate, will show how much of the Q1 beat was migration and how much was durable demand.
Sources: SAP Quarterly Statement Q1 2026 (SAP / PR Newswire), Benzinga — SAP Q1 report details, Constellation Research — DSAG 2026 Investment Report, The Stack — “Subscription pricing is foolish, says SAP CEO”, Innobu — SAP Joule 2026.