The Backlog Finally Turns Up, the Margin Turns Down, and the Buyback Sits Out the Low

Results vs. Consensus

Consensus for SAP is quoted by US desks in dollars against the NYSE ADR, and for this quarter the dollar EPS estimate is struck on the IFRS line rather than the non-IFRS line. Figures below are translated at the 1.1378 USD/EUR spot prevailing on the day of the release, which reconciles the dollar-quoted actuals to the euro income statement exactly.

Year-over-Year Comparison

First Half 2026

The half-year view is the fairer frame for margin, and it is materially better than the quarter. Non-IFRS operating margin for the first half expanded 100bp reported and 120bp at constant currency to 28.9%. The CFO leaned on precisely this point when challenged, and he is right that a single quarter carrying an unusual concentration of items is a poor basis for extrapolation. It is also true that the concentration ran the other way in the first quarter, when a share-price-driven credit flattered the result.

Quality of Beat/Miss

• Revenue: Cloud growth decelerating to 24% at constant currency from 27% was pre-announced at the first quarter and arrived as promised. Management added a second reason beyond the Q1 one-offs: a base effect, since Q2 2025 ran roughly two points above Q1 2025. Both explanations are verifiable and neither implies demand deterioration. Total revenue at 11% constant currency held the full-year 2025 rate exactly. Software support fell 7% at constant currency, an acceleration from the 6% decline in the first quarter and consistent with the guided steepening.
• Margins: This is the weak point and it deserves to be stated without softening. Non-IFRS operating margin contracted for the first time in the four quarters we have covered, and cloud gross margin fell 70bp at constant currency. Management identified five causes: slower cloud and total revenue growth, the unusually low first-quarter share-based compensation expense creating a difficult sequential comparison, accelerated research and development investment, higher marketing around the Autonomous Enterprise launch, and Reltio dilution from its May 7 close. Four of those five are timing or deliberate reinvestment. The cloud gross margin decline is the one that is neither, and management attributed it to one-time test-environment investment for the new platform plus the cost of delivering end-to-end sovereign environments.
• EPS: Non-IFRS EPS of €1.59 grew 6%, the slowest of the four quarters, tracking the 7% reported operating profit growth with no below-the-line help. The non-IFRS effective tax rate of 30.8% ran above the roughly 29% full-year guide, which cost roughly two points of EPS growth. The 30% IFRS EPS growth is not comparable, resting on tax-exempt income excluded from the non-IFRS figure.
• Cash: The standout line. Free cash flow grew 27% to €3,002M in a quarter when operating profit grew 7%, and operating cash flow grew 22%. First-half free cash flow of €6,250M is up 5% even though the current first half absorbed the €408M Teradata settlement payment. Adjusting for that, underlying first-half free cash flow growth is closer to 12%. Cash conversion is running well ahead of accounting profit, which is the correct direction for a business whose value rests on the free cash flow yield.

Segment Performance

Current Cloud Backlog — The Turn

Current cloud backlog reached €22,929M, up 27% reported and 26% at constant currency. This is the first sequential acceleration in the metric across the four quarters we have covered, and it reverses a pattern in which backlog growth trailed cloud revenue growth. The CEO framed it as exactly that.

"Current cloud backlog grew 26%, an acceleration compared to Q1. And after two quarters where our CCB was lagging behind cloud revenue growth, it was a welcome trend reversal in this important forward-looking indicator, a great result, especially given the volatile environment." — Christian Klein, CEO

Management disclosed the inorganic component without being asked, and quantified it.

"Current cloud backlog reached almost EUR 23 billion, up 26%, benefiting from the first-time inclusion of Reltio, which only contributed less than 1 percentage point to the constant currency growth rate." — Dominik Asam, CFO

Assessment: This is the most important datapoint in the quarter and it directly addresses the metric that caused the January collapse. Even stripping the sub-one-point Reltio contribution, organic backlog growth held at or slightly above the prior quarter, against a guide that called for deceleration and against a macro backdrop management describes as actively weighing on decision-making. Management continues to guide for slight deceleration exiting the year, which after this print looks conservative rather than ominous.

Cloud ERP Suite — Deceleration From a High Base

Cloud ERP Suite grew 27% at constant currency to €5,525M, down from 30% in the first quarter, and now represents 88% of cloud revenue. Two mechanical effects explain the step down: the first-quarter one-offs management flagged in April, and a base effect from a strong second quarter of 2025.

"As expected, the year-over-year cloud revenue growth rate declined sequentially to 24%, reflecting several quarter-specific effects that particularly benefited the 2026 year-on-year comparison in the preceding first quarter… Also recall that in Q2 2025, we had roughly 2 percentage points higher cloud revenue growth than in Q1. So there is a strong basis effect here." — Dominik Asam, CFO

Assessment: A 27% constant-currency growth rate on a €5.5B quarterly base is not a deceleration worth worrying about, and it remains above the 23 to 25% full-year cloud guide. The genuine watch item is the concentration: at 88% of cloud revenue, Cloud ERP Suite is now effectively the whole cloud line, so its rate and the headline rate are the same number. There is no longer any diversification inside cloud to cushion a step down.

Cloud Gross Margin — The Stall Continues

Non-IFRS cloud gross margin fell 70bp at constant currency to 74.6%, following a 10bp decline in the first quarter. Across four quarters the level has oscillated in a narrow 74.6 to 75.2% band, so this is more accurately described as expansion having stopped than as a decline having begun. That distinction matters, because the thesis rests on cloud mix being accretive, and it is accretive at a stable 75% even if it stops improving.

"Now in Q2, we had some onetime investments into the test environments for our new platform into Joule Work. So we had to do that. And we have a cost increase for the sovereignty environments we are delivering end-to-end." — Christian Klein, CEO

Management also offered a forward argument specific to AI cost of goods, pointing to model choice as a lever it now controls.

"I mean there is nothing which now stops us with the ability to switch models to always choose the best model for an ideal price outcome. And we see already that for many agents we are now developing, especially for Joule Work, we don't always need to use the expensive frontier models." — Christian Klein, CEO

Assessment: The sovereignty cost point is the one with legs. Delivering end-to-end sovereign environments is structurally more expensive than running on hyperscaler infrastructure, and management has been telling us for three quarters that sovereign demand is a growth driver. A margin cost attached to the fastest-growing part of the pipeline is not a one-time item. The model-routing argument is credible and unproven: it is the right architecture, and no evidence was offered that it has yet reduced cost of goods. We now model cloud gross margin flat at roughly 74.5 to 75% rather than expanding.

Support and Services — Both Declining, Only One by Choice

Support revenue fell 7% at constant currency to €2,439M, a steepening from the first quarter's 6% and consistent with the guided acceleration. Services fell 2% at constant currency, and the CEO tied the decline explicitly to redeployment rather than lost demand.

"This comes despite a slight decline in services revenue as we reallocate consultants to build agents and foster AI adoption." — Christian Klein, CEO

Assessment: Consistent with the framing given in April and now visible for a second quarter, which makes it a policy rather than a one-off. The combined drag from support and services is roughly 35% of revenue declining at low-to-mid single digits, which is exactly why total revenue grows 11% while cloud grows 24%. Nothing here changed; the arithmetic is simply now fully disclosed on every line.

Regional Performance

Cloud revenue performance was again particularly strong in APJ and EMEA and solid in the Americas, with Brazil, France, Germany, Italy, India, South Korea and Spain outstanding, and Australia, Singapore and the United States particularly strong.

Assessment: The Americas has now carried the "solid" descriptor for four consecutive quarters against APJ and EMEA's "particularly strong," which is a settled pattern rather than a signal. Saudi Arabia remains off the outstanding list for a second quarter, the clearest regional trace of the Middle East disruption.

Key Topics & Management Commentary

Overall Management Tone: Confident on demand and defensive on cost, which is the reverse of the posture through most of the past year. The backlog reacceleration gave management a genuine win to lead with and it did, at length. On margin the response was to reframe the measurement window from the quarter to the half and then to the full year, which is analytically fair and was also repeated often enough to read as prepared. The most substantive disclosure came in the Q&A rather than the prepared remarks: management explicitly considered raising the organic profit guide to absorb acquisition dilution and decided against it in favor of investment room.

1. The Guide Cut That Is Not a Guide Cut

The full-year non-IFRS operating profit outlook moved to €11.8–12.2B at constant currency from €11.9–12.3B, growth of 13 to 17% against the previous 14 to 18%. Every other line was maintained: cloud revenue at €25.8–26.2B, cloud and software at €36.3–36.8B, free cash flow at approximately €10B, and the tax rate at approximately 29%. The reason given is dilution from the Dremio and Prior Labs acquisitions, which closed in July, projected in excess of €100M.

Pressed on it, the CFO disclosed the decision behind the decision.

"Now — it is true though that the M&A acquisitions we did, they will have actually an impact which is in excess of EUR 100 million. And we, of course, had a debate now should we basically upgrade the underlying organic guidance and say we absorb that on top or should we keep the bigger room to drive aggressive growth and we decided to go for the latter." — Dominik Asam, CFO

He then restated the organic position directly.

"To cut it short, we are fully on track on our initial non-IFRS operating profit outlook, which obviously did not include any M&A effects despite all macro headwinds so far this year." — Dominik Asam, CFO

Assessment: This reframes the quarter's most negative headline. The organic profit outlook set in January is intact after two quarters of Middle East disruption, and on management's account was strong enough that absorbing €100M of acquisition dilution within the original range was a live option. Choosing investment room over an optical raise is defensible and it is also convenient, because it converts a discretionary spending decision into a guidance revision the market must take on trust. The check is the 80 to 90% expense-to-revenue-growth ratio, which management reaffirmed for 2027 including the acquisition J-curves.

2. The Margin Contraction, Itemized

Non-IFRS operating margin fell 70bp to 27.8%, and management gave a five-part attribution rather than a single cause.

"The quarter-on-quarter deceleration in IFRS and non-IFRS operating profit growth is mainly caused by lower cloud and total revenue growth in Q2 as compared to Q1 2026, an unusually low stock-based compensation expense in the first quarter of 2026, accelerated investments into research and development, higher marketing expenses in the quarter of the launch of the Autonomous Enterprise and finally, the slightly dilutive impact of the Reltio acquisition, which just closed on May 7." — Dominik Asam, CFO

The CFO's framing was that a single quarter with this many overlapping items is not extrapolable.

"So this is why I mentioned in my introductory remarks, from my perspective, there is no point in extrapolating just one single quarter where we had a concentration of all the factors I've already mentioned, which I can also reiterate if you're interested in. It was a little bit of a special quarter, so don't read too much into that." — Dominik Asam, CFO

Assessment: The itemization confirms the analysis we published in April, when we flagged that the first quarter's €135M share-based compensation credit was a share-price artifact that would reverse and create a comparison problem. Management now cites the "unusually low stock-based compensation expense in the first quarter" as a cause of the second-quarter deceleration, which is the same point from the other side. Taking the half rather than either quarter, non-IFRS operating margin expanded 120bp at constant currency to 28.9%, which is a good outcome. The residual concern is that the margin story has moved from delivered to promised, and the promise now sits in the second half.

3. The Second Half Has to Do Two Opposite Things

The guidance arithmetic sets up a demanding and internally awkward second half, and one questioner did the work.

The CFO's answer to how that happens was operational rather than structural.

"We are frankly optimizing massively now on how we spend tokens by virtue of a very tight controlling. We can now, on a very granular basis, see who's using what tool and what is output driven here, and we will funnel the tokens in a way that gives us a better bang for the buck. And there are also measures on cost containment to really make sure that we focus our resources where it really matters." — Dominik Asam, CFO

Assessment: This is the single most demanding element of the guidance package and the answer is thinner than the question deserved. Token-spend optimization and cost containment are real levers but they are not obviously worth five points of profit growth acceleration. The mitigating factors are that the second-quarter comparison base was distorted by the launch marketing and the first-quarter compensation credit, both of which lapse, and that the second half faces an easier operating profit comparison. We are trimming our full-year estimate toward the lower half of the guided range rather than the midpoint.

4. No Shares Repurchased in the Second Quarter

The buyback disclosure reads, verbatim in substance, the same as it did three months ago: 16,280,097 shares repurchased at an average price of €161.16 for approximately €2.6B, now stated as of June 30 rather than April 1. The number is unchanged. SAP bought nothing in the second quarter.

Over that same quarter the shares made new lows, closing the period at $146.38, which translates to roughly €129 and is approximately 20% below the €161.16 average paid in the first tranche. The programme has €7.4B of its €10B authorization remaining and runs to the end of 2027.

Assessment: This is the worst fact in the quarter and it drew no questions. In January management characterized the €10B authorization as reflecting confidence in the sustainable strength of the business. Deploying €2.6B at €161.16 and then standing aside while the stock falls to €129 is the opposite of what a price-sensitive repurchase programme should do, and it happened in the same quarter SAP closed three acquisitions. A company that finds capital for Reltio, Dremio and Prior Labs but not for its own equity at a five-year low is making an implicit valuation statement about both. The charitable reading is that tranche mechanics and acquisition funding constrained the window. The charitable reading was not offered, because nobody asked.

5. The European Maintenance Investigation Resolves Without a Fine

The regulatory overhang we have flagged in every report since initiation, and which had drawn zero analyst questions in three consecutive quarters, was finally raised and finally resolved. The outcome is a commitments agreement rather than a penalty.

"Now you said EU ruling, this is an agreement between the EU and SAP to commit to certain mitigants in terms of flexibility on maintenance… And we have actually also in the past, granted flexibility and opportunities for these customers to adjust their software spend to the needs. And that has been, to some degree, now formalized in this agreement. And in some degrees, some more flexibility has been granted in limited scenarios for the customer." — Dominik Asam, CFO

Management sized the risk as manageable and made a point about the underlying economics that had not previously been disclosed.

"So, and the last point I want to mention, we do actually see quite a nice pickup in returns from third-party maintenance. that gives us also confidence when people try that for a while, they tend to be more nervous over time about incidents happening and then come back." — Dominik Asam, CFO

Assessment: A materially better outcome than the tail risk we had been carrying. Our concern was never the fine; it was that a remedy constraining maintenance policy would blunt the economic pressure driving cloud conversion. A commitments agreement that formalizes flexibility SAP already granted, in limited scenarios, does not do that. The customers-returning-from-third-party-maintenance disclosure is genuinely new and cuts against the most credible bear argument on the support base. We are downgrading this risk in our thesis accordingly. The residual point management made stands on its own: extended maintenance for the legacy estate expires in 2030 regardless, so any flexibility granted now has a fixed end date.

6. Sapphire and the Autonomous Enterprise Launch

The Sapphire conference in Orlando, which management had spent the April call deferring to, delivered the Autonomous Enterprise launch. The new business AI platform ships this quarter across three pillars: build, via Joule Studio with a choice of Anthropic, Cohere, Google, Mistral AI, OpenAI and open-weight models; context and reason, combining Business Data Cloud with the newly acquired Dremio and Reltio technologies and an ontology layer; and run and govern, covering agent lifecycle management and compliance across more than 130 countries.

The delivery commitments are specific and dated: close to 50 assistants by the end of the third quarter, more than 400 Autonomous Suite agents by the end of the year, three additional ERP migration assistants with ten underlying agents this quarter, and the Joule Work interface launching in the third quarter across a stated 350 million end users.

"After Sapphire, the beta programs for our new platform, Suite and Joule Work were immediately oversubscribed and initial customer feedback has been excellent. This makes us very confident about the successful launch in Q3." — Christian Klein, CEO

Assessment: Management delivered what it promised in April and attached dates to it, which is more than most enterprise software AI narratives offer. The commercial model is the part still missing: no pricing, no attach expectation, no revenue contribution framing. Oversubscribed beta programmes are a real signal of interest and a weak signal of willingness to pay. The 400-agent target also sits oddly against the CEO's own dismissal of agent-count marketing two quarters ago, when he asked which salesperson could sell 800 technical agents.

7. Three Acquisitions in Six Months

SAP has now closed Reltio on May 7, plus Dremio and Prior Labs in July. Each has a stated technical rationale within the platform architecture: Reltio governs master data models, Dremio brings Apache Iceberg-native technology allowing SAP and non-SAP data to be analyzed together without copying, and Prior Labs enables agents to generate tabular predictions.

The CEO was explicit that none was bought for revenue.

"I mean, all three acquisitions now we did was not to acquire growth. I mean, first of all, all three acquisitions have only a very, very minor impact on CCB and revenue. The latest two actually have no impact on cloud revenue." — Christian Klein, CEO

The CFO quantified the drag precisely: Reltio produced an IFRS loss after tax of €10M since closing, reducing to a high single-digit million euro amount on a non-IFRS operating profit basis, while Dremio and Prior Labs together will weigh on the second half by a very low triple-digit million euro amount.

Assessment: The technical logic is coherent and each acquisition maps to a named gap in the platform the CEO described. The tension is with the capital allocation record: management stated in April that tuck-ins would be "clearly not meant to acquire revenue," and it has been consistent with that, but it has now spent on three of them while its own shares trade at a five-year low and the buyback sits idle. The Prior Labs rationale is the most interesting, because it targets tabular prediction, and as the CFO noted SAP holds one of the largest reservoirs of tabular business data in existence. If any of these purchases proves strategically decisive, it is that one.

8. Research and Development Steps Up Hard

A question surfaced the disparity between research and development headcount growth of roughly 3% year over year and cost growth of roughly 14%, against 2% and 2% respectively in the first quarter. Management framed the increase as a reshuffle with a defined duration.

"Absolutely the latter one. I mean, give us some time, please. I mean, look, I mean, this is the second transformation we are now in. And as I mentioned before, first, you need to reshuffle the backlog. I mean our backlog was full of feature requests from our customers, and we need to have the time to reshuffle this now to AI." — Christian Klein, CEO

The CFO placed it inside the existing envelope rather than outside it.

"And recall, we always gave that kind of formula to say that total expenses will grow at 90% of the revenue growth. And we always said we want to have some wiggling room within any specific line item." — Dominik Asam, CFO

Assessment: A 14% cost increase on 3% headcount growth is compensation mix, not hiring: SAP is paying up for scarce AI and data engineering talent, which is consistent with the "quality not quantity" posture management has held since October. The line-item flexibility argument is legitimate as long as the aggregate envelope holds, and the aggregate envelope is the thing to grade. Management explicitly said the point will sit toward the lower end of the 80 to 90% range this year, which is a small admission that the envelope is being used rather than beaten.

9. AI Pricing: The Argument Management Wants to Make

Late in the call the CEO reached for the argument that underpins the entire long-term case, and it is worth isolating because it is the first time he has framed AI as a pricing reset rather than a product feature.

"We have thought that we're going to achieve such a gross margin, by the way, five years ago. Now with AI, I mean, now we can completely reset the price level." — Christian Klein, CEO

Assessment: This is the bull case in one sentence and it is entirely unevidenced so far. If SAP can price AI-infused workflows against business outcomes rather than against seats, the revenue-per-customer ceiling rises materially and the consumption transition becomes accretive rather than dilutive. Nothing in this quarter's numbers demonstrates it, and management offered no pricing data. Treat it as an option with unknown strike rather than as guidance. It is also the natural subject for the commercial model announcements that have now been promised twice.

10. Free Cash Flow Runs Ahead of Profit

Free cash flow grew 27% to €3,002M in a quarter when non-IFRS operating profit grew 7%, and operating cash flow grew 22%. First-half free cash flow of €6,250M is up 5% despite absorbing the €408M Teradata settlement outflow in the first quarter.

Assessment: Against the approximately €10B full-year guide, the second half requires roughly €3.75B versus €2.30B in the same period last year, a 63% increase. That looks demanding until the composition is examined: the second half of 2025 carried the tail of the restructuring payouts and a fourth quarter with an unusually low free cash flow contribution. The guide is achievable but not comfortable, and free cash flow is the metric on which our valuation case principally rests, so it is where we would look first for a miss.

Guidance & Outlook

Disclosed Currency Translation Impact

The sensitivity table assumes 1.14 USD/EUR against a 1.13 constant-currency planning rate. The notable change is that currency becomes a tailwind in the third quarter for the first time in this cycle, adding 1.5 percentage points to reported cloud revenue growth and 1.0 point to cloud and software.

Total revenue growth is now a number. The outlook quantifies the 2025 base at 10.6% and guides 2026 to a similar level. This closes the last piece of the ambiguity we flagged in January, when the guide said "accelerate through 2027" while the arithmetic implied flat. The language was corrected in April; the figure has now been supplied.

Street at: Consensus entering the print had total revenue in line at approximately €9.88B and IFRS EPS around €1.76. Both were met or beaten. Current cloud backlog growth of 26% at constant currency was above expectations, which had been anchored on management's own guidance for deceleration. The operating profit trim was not anticipated.

Analyst Q&A Highlights

Whether the Beat-and-Raise Pattern Has Ended

The opening question named the pattern investors had grown used to and asked whether the second-quarter margin and the full-year trim signal less room to manoeuvre than assumed. The response drew the measurement window wider and then disclosed the decision behind the trim.

Q: "I think investors have got used to the kind of beat and raise cycle from SAP on the operating income line and just maybe a little bit of a surprise to see the weaker second quarter. I appreciate there are good reasons behind that. But maybe also the cut for the full year suggests there was a little bit less room for maneuver than people thought… could you just talk a little bit about going into a bit more depth there around how you think about how much room there is for margin expansion in the business over the next 12 to 24 months?"
— Adam Wood, Morgan Stanley

A: "For instance, if you look at the first half, you see that our operating leverage envelope has been well respected. And we will also see — we also see that for the full year '26, we'll be well in that kind of operating leverage formula. And that's even including the acquisitions, all the acquisitions we have made. And so of course, it pushes the kind of point a little bit more to the downside of the range. But we think the real absolute highest priority is to drive the AI transformation forcefully and to protect the top line in the current environment."
— Dominik Asam, CFO

Assessment: The admission worth extracting is "it pushes the kind of point a little bit more to the downside of the range." Management is telling the market it will land at the less profitable end of its own operating leverage commitment, and framing that as a deliberate choice to fund the AI transition and defend growth. That is an honest trade to disclose and it is also the definition of a margin story going on hold. The half-year framing is legitimate; the first half did expand margin 120bp at constant currency.

Reconciling a Reaccelerating Backlog With a Decelerating Revenue Guide

The sharpest question of the call observed that the first half is running ahead of the cloud guide while backlog is reaccelerating, and asked how much of the implied second-half deceleration is genuine expectation rather than prudence.

Q: "H1 is running slightly ahead of your guidance. So we're looking at quite a material deceleration in cloud revenue growth in the second half to get to your guidance midpoint. Again, this despite adding more M&A despite CCB growth reaccelerating in Q2 and also growing ahead of cloud revenues, which is usually positive directionally for trailing cloud revenue growth. So I guess how much of this is just prudence given what's going on in the world, but how much is actually what you really expect?"
— Ben Castillo-Bernaus, BNP Paribas

A: "I mean because of the macro now in Q2, definitely, I mean, look at the CCB, super strong. We don't see any major backlog because of macro. I mean, obviously, in the Middle East, a few deals here and there got delayed, but definitely not at the broader scale… obviously, now with being seven months in the year, as I just said, of course, the predictability becomes better and better. So we are more and more confident that we can also hit the guidance what we outlined at the beginning of the year despite all of the volatility we see out there in the market."
— Christian Klein, CEO

Assessment: The answer confirms prudence without conceding it in words. "More and more confident that we can also hit the guidance" from a first half running above the range is the language of a company that expects to land at or above the midpoint and does not wish to say so with two quarters of geopolitical uncertainty remaining. Combined with the backlog reacceleration and the confirmation that macro-driven delays were narrow rather than broad, the cloud revenue guide now looks conservative. That is the single most constructive inference available from this call.

Whether the Second-Half Profit Reacceleration Is Achievable

A question laid out the guidance arithmetic explicitly, noting that first-half operating profit growth of 16% at constant currency against a full-year midpoint implies a second half well above the 9% delivered in the second quarter.

Q: "EBIT slowed to 9% in Q2 for the reasons that you laid out, but you had a strong Q1. So overall H1 EBIT growth constant currency was 16%. And the guidance midpoint implies sort of mid-teens or so EBIT growth in the second half, which would imply a reacceleration relative to the Q2 EBIT growth rate. Can you help us understand what would drive that reacceleration from the Q2 EBIT growth rate?"
— Toby Ogg, JPMorgan

A: "I did mention in my introductory comments that Q2 was a little bit of an abnormal situation… We are frankly optimizing massively now on how we spend tokens by virtue of a very tight controlling… And there are also measures on cost containment to really make sure that we focus our resources where it really matters. So — but you nicely summarized the thinking we have in terms of H1, H2. This is exactly what we're going to do."
— Dominik Asam, CFO

Assessment: Management accepted the questioner's arithmetic without amending it, which makes the second-half profit reacceleration an explicit commitment rather than an inference. The stated levers, token-spend optimization and cost containment, are thin relative to a five-point acceleration, and the more likely mechanism is simply that the second-quarter launch marketing and the first-quarter compensation distortion both lapse. This is the number to grade in October, and it is the reason we model toward the lower half of the guided range.

The European Maintenance Agreement and Whether It Slows Migration

A question raised the settlement with the European Commission and the concern that granting customers more maintenance flexibility could extend the life of the legacy estate. Management addressed both the substance and the underlying economics.

Q: "Can you just say something on the recent European ruling on maintenance? There's some suggestion, I think, that giving customers more maintenance options potentially opens the door to them staying on ECC for longer. Do you think this ruling actually changes any behavior on the ground?"
— Charles Brennan, Jefferies

A: "And for some customers that are actually prioritizing spend, lower spend over the advantages of maintenance and support I described, there might be an impact, but we think we can manage it. And don't forget that one is also phasing out more and more as we convert the customers onto cloud and RISE. And that has nothing to do actually with ECC versus S/4 transition because at some point in time, that maintenance will anyhow expire. In 2030, ECC maintenance is basically zero anyhow."
— Dominik Asam, CFO

Assessment: The decisive point is the 2030 expiry. Whatever flexibility the agreement grants operates inside a window that closes on a fixed date, so it can change the timing of migration but not the requirement. Management conceded a manageable revenue impact from spend-optimizing customers rather than denying one, which is the credible answer. Combined with the disclosure that customers are returning from third-party maintenance providers, the regulatory tail risk we have carried since initiation is materially reduced.

The Logic of Three Acquisitions in Six Months

A question asked management to walk through the reasoning behind the two most recent acquisitions specifically. The technical answer came from the CEO; the more interesting observation came from the CFO.

Q: "Christian, can you just follow up on that last point on the most two recent acquisitions you'done. I'd be curious sort of if you could just go over again the thought process behind specifically Prior Labs…"
— S. Kirk Materne, Evercore ISI

A: "Maybe one — maybe at the risk of stating obvious, what is also interesting about Prior Labs is that it's trained on tabular databases. I mean, no secret, SAP has probably the biggest reservoir of data of tabular…"
— Dominik Asam, CFO

Assessment: The tabular-data point is the most strategically loaded remark on the call. General-purpose language models are trained on text; SAP's asset is structured transactional data at a scale no model provider holds. A foundation model purpose-built for tabular prediction, applied to that reservoir, is the one place SAP could hold a genuinely non-replicable technical advantage rather than a domain-knowledge one. It is also unproven, unpriced and unquantified, and it was raised as an aside rather than as a strategy.

Research and Development Cost Growth of 14% on 3% Headcount

A question isolated the divergence between research and development headcount and cost growth, and asked whether the shift from sales and marketing into research and development is structural.

Q: "A question on the R&D side of things. So headcount looks to be up 3% year-on-year and the costs 14%. In Q1, the numbers were 2% and 2%."
— Michael Briest, UBS

A: "I mean, first of all, in the last 12 months, indeed, we still invested into new job profiles in R&D, data scientists, data engineers, we invested into full stack developers for industry…"
— Christian Klein, CEO

Assessment: A four-fold gap between cost and headcount growth is a compensation-mix disclosure, and it confirms what management has been describing qualitatively for three quarters: SAP is trading headcount for talent density in AI and data roles. The strategy is right and it is expensive in the near term, which is precisely the margin pressure visible this quarter. The commitment that follow-on years show "very healthy R&D ratios" is the forward promise attached, and it belongs on the watch list rather than in the model.

What They're NOT Saying

1. Why no shares were repurchased in the second quarter: The buyback disclosure is numerically identical to April's while the stock fell roughly 20% below the average price already paid. Three acquisitions closed in the same window. No explanation was offered and no question was asked. This is the largest unexamined item in the release.
2. Any commercial model for the Autonomous Enterprise: Close to 50 assistants by the end of the third quarter, more than 400 agents by year-end, Joule Work across 350 million end users, and beta programmes oversubscribed. No pricing, no attach target, no revenue contribution, no margin profile. Commercial model detail has now been promised at two consecutive events.
3. Whether the sovereign cost drag is permanent: Cloud gross margin fell 70bp at constant currency and management named end-to-end sovereign environments as a cause alongside one-time platform test investment. Sovereign demand is described as a structural growth driver, so a cost attached to it is structural too. The two statements were made minutes apart and never reconciled.
4. How token costs scale with agent deployment: Management described optimizing token spend as a second-half profit lever and model routing as a gross margin lever, without disclosing what AI inference currently costs, what it is expected to cost at 400 agents, or how it is recovered in pricing.
5. The second-half free cash flow bridge: Approximately €3.75B is required against €2.30B in the same period last year. The full-year guide was reaffirmed without a bridge, and free cash flow drew no questions despite being the metric the CFO built an explicit formula around in January.
6. Whether more acquisitions are coming: Three closed in six months and the CEO described an ongoing search for gaps in the data and AI space. Given that the current-year profit guide has now been adjusted twice around acquisition dilution, whether a fourth would trigger a third adjustment was not addressed.
7. Customer NPS: Missed its target range for 2025 and has not been updated or discussed in either of the two subsequent quarters.

Market Reaction

• Pre-print setup: The NYSE-listed shares closed at $146.38 on July 23, which was the 52-week closing low. The stock entered the print down 39.7% year to date against the S&P 500's +8.2%, down 49.6% over the trailing twelve months and down 4.6% over the trailing thirty days. The 52-week closing range was $146.38 to $306.29.
• Next-day session: Shares gapped up to open at $153.75, a 5.0% advance, traded between $153.21 and $161.31, and closed at $160.00, up 9.3% or $13.62.
• Volume: 6.2M shares against a 2.7M thirty-day average, roughly 2.3 times normal.
• Benchmark: The S&P 500 was unchanged on the session, so the entire move was company-specific.

The reaction was driven by one number. Current cloud backlog growth of 26% at constant currency reversed two quarters of underperformance on the metric that caused January's 15% collapse, and it did so against management's own guidance for deceleration. Everything else in the release was mixed to negative: margin contracted, the profit guide was trimmed, and non-IFRS EPS grew 6%. The market bought the leading indicator and ignored the rest, which is a reasonable ordering for a stock whose de-rating has been entirely about forward growth.

The setup amplified it. A 9.3% single-session gain recovers roughly one-seventh of the year-to-date decline and leaves the shares 48% below their 52-week high. Entering a print at a 52-week low with the market up 8% on the year means positioning is one-sided and the bar for a positive surprise is close to the floor. The close at $160.00, near the upper end of the day's range, on 2.3 times normal volume, is the third consecutive quarter in which the reaction session has finished strong.

What has not happened across those three quarters is a durable recovery. The shares closed at $200.21 after the January print, $175.26 after April's, and $160.00 after this one. Each print has produced a positive or resilient session, and the stock has been lower at each subsequent print. The de-rating is not being driven by the quarterly results.

Street Perspective

Debate: Is the Margin Stall Cyclical or Structural?

Bull view: First-half non-IFRS operating margin expanded 120bp at constant currency to 28.9%. The second-quarter contraction is a concentration of identifiable items, most of which lapse: launch marketing, an unusually low first-quarter compensation base, acquisition dilution, and platform test-environment build costs. Management reaffirmed the 80 to 90% expense-to-revenue-growth envelope including acquisition J-curves, and free cash flow grew 27% in the same quarter.

Bear view: Cloud gross margin has now failed to expand for two consecutive quarters and the named cause, sovereign delivery, is attached to the fastest-growing part of the pipeline rather than to a one-off. Research and development costs grew 14% on 3% headcount growth. Management concedes it will land at the lower end of its own operating leverage range. A business that stops converting mix shift into gross margin has lost the mechanism that made the cloud transition worth funding.

Our take: The bear case has identified the right variable and overstated its trajectory. Cloud gross margin has oscillated between 74.6 and 75.2% across four quarters rather than declining; what has stopped is expansion, not the level. At a stable 75% the mix shift remains accretive, because cloud gross margin still sits above the blended corporate rate of 73.4%. We now model cloud gross margin flat rather than expanding, which removes a source of upside from our estimates without breaking the thesis. The genuine risk is sovereign scaling: if sovereign becomes a large share of a growing cloud base and carries a persistent cost premium, the flat assumption becomes optimistic.

Debate: Does the Backlog Reacceleration Mark the Bottom?

Bull view: Backlog growth accelerated to 26% at constant currency, with less than one point from Reltio, against a guide for deceleration and in a quarter management describes as macro-impaired. Cloud revenue for the first half is running above the full-year guide. Management characterized macro-driven deal delays as narrow rather than broad. The metric that broke the stock in January has now recovered.

Bear view: One quarter does not reverse a trend, management still guides deceleration exiting the year, and the range of outcomes remains "wider than we would like" on management's own description. The second half carries the bulk of annual bookings and the Middle East situation is unresolved. Backlog also accelerated partly because the prior-year comparison was soft.

Our take: The bull case is stronger and the market agreed, awarding 9.3% for essentially this one fact. The reason to weight it heavily is that it is the first datapoint in four quarters that contradicts rather than confirms the deceleration narrative, and it arrived under conditions that should have produced the opposite. What it does not resolve is the 2027 acceleration thesis, which needs backlog to convert into total revenue growth, and that remains untestable until next year. The correct reading is that the downside case has weakened, not that the upside case has been proven.

Debate: Is the Capital Allocation Defensible?

Bull view: Each acquisition maps to a specific named gap in the platform architecture, none was bought for revenue, and the combined dilution is roughly €100M against €12B of operating profit. Prior Labs in particular targets tabular prediction, where SAP's data reservoir is unmatched. The buyback has €7.4B of authorization remaining and runs to the end of 2027, so nothing has been forfeited.

Bear view: Management deployed €2.6B of buyback at €161.16 and then bought nothing as the shares fell roughly 20% below that level, while finding capital for three acquisitions in the same window. That is the reverse of price-sensitive capital allocation, and it happened without explanation or challenge. The guide has now been adjusted twice around acquisition dilution in six months.

Our take: The bear case is correct and it is the least defensible element of this quarter. A repurchase programme announced as a statement of confidence, executed heavily at a higher price and then paused at the low, is either poorly administered or a signal management does not intend to send. Either reading is negative. The acquisitions themselves are individually defensible and cheap relative to the balance sheet; the issue is the relative allocation, not the absolute amount. We expect this to be asked in October and we would treat a resumed buyback in the third quarter as a meaningful positive.

Model Update Needed

Valuation: At the reaction-day close of $160.00 and the FMP-verified 1.1371 USD/EUR spot on July 24, the implied price is approximately €141 per ordinary share. Share count of approximately 1,150M derives from second-quarter non-IFRS profit after tax of €1,828M against basic non-IFRS EPS of €1.59. Market capitalization is therefore near €162B.

On those figures SAP trades at approximately 23x trailing FY25 non-IFRS EPS of €6.15 and approximately 20x our FY26 estimate of €7.05 to €7.20. Free cash flow yield on the €9.5 to 10B range is approximately 5.9 to 6.2%. The unexecuted €7.4B of buyback authorization is equal to roughly 4.6% of market capitalization.

The four-quarter arc, stated plainly: we initiated at Hold in October at roughly 39x forward earnings on a 2.9% free cash flow yield; upgraded in January at 23x and 5.1%; maintained in April at 21x and 5.8%; and the shares now trade at roughly 20x and 6.2%. Across those four quarters SAP delivered FY25 above both profitability guides, grew non-IFRS EPS 36% in 2025 and 12% in the first half of 2026, reaccelerated its leading indicator, and resolved its regulatory overhang. The share price fell 42% from our initiation. Our earnings estimates have risen or held in every period and the multiple has compressed by roughly half. The upgrade has been wrong on price for two consecutive quarters and we are not going to describe that as anything else.

Thesis Scorecard Post-Earnings

Grading Last Quarter's Commitments

Overall: Thesis net stronger on the forward-looking pillars and weaker on margin. Two bear points improved, one resolved outright, one new bear point added on capital allocation, and the margin pillar remains at risk. Six of nine prior-quarter commitments were delivered or called correctly.

Action: Maintaining Outperform. The case rests on three things that are now observable rather than promised: a leading indicator that has turned up, free cash flow compounding at 27% in the quarter and guided to approximately €10B, and a valuation of roughly 20x forward earnings on a 6.2% free cash flow yield for a business growing cloud revenue at 24% at constant currency. The regulatory tail risk we carried from initiation has resolved without a fine. What has deteriorated is margin, and management has told the market it will run at the less profitable end of its own leverage range while it funds the AI transition.

What would change our mind: we would downgrade to Hold on a third consecutive quarter without cloud gross margin expansion combined with a second-half operating profit outcome below the guided range, which together would convert a deliberate investment cycle into a structural margin reset. We would downgrade to Underperform if backlog growth falls below 22% at constant currency while margin continues to compress. Conversely, a resumed buyback, a disclosed commercial model for the Autonomous Enterprise, and second-half operating profit growth reaching the guided mid-teens would together justify raising conviction rather than merely maintaining the rating.