SAP: How a Legacy Software Company Became Europe’s Most Valuable

SAP is the only European technology company of genuine global scale, and its recent success came from a transition that most incumbents fail. Converting a licence business into a subscription business destroys reported revenue before it rebuilds it, and the companies that survive it are those whose customers have nowhere else to go. This case study opens the software pillar of the Germany Company Stories hub.

Why is a licence-to-subscription transition so dangerous?

Because the accounting works against you for several years. A perpetual licence recognises a large payment immediately; a subscription spreads an equivalent commitment across the contract term. Converting customers therefore reduces reported revenue even when the underlying business is growing.

The cash effect is similar. Upfront licence payments become monthly or annual instalments, which weakens near-term cash flow while the cost base, principally research and sales, continues at full weight.

Most incumbents attempt to manage this by transitioning slowly, which produces the worst outcome: a decade of mixed models, confused sales incentives, two product lines to maintain and a customer base uncertain about which platform will receive investment.

SAP's approach was to commit publicly and accept the margin compression, including a period in which investors punished the shares for the same decision they later rewarded. That sequence, near-term pain for structural improvement, is only survivable with a defensible installed base.

Why can enterprise customers not simply switch?

Because the software is entangled with the business. An enterprise resource planning system holds the master data, the process logic, the financial close, the compliance controls and typically two decades of customisation, and replacing it means re-engineering how the company operates.

Migration projects at large enterprises routinely run for years and cost multiples of the licence value in consulting fees. The risk is operational rather than financial: a failed migration can prevent a company from invoicing customers or closing its books.

That produces the strongest switching cost structure in commercial software, and it is why the vendor can raise prices, change the commercial model and impose migration deadlines with more confidence than almost any other supplier.

The limit is patience. Switching costs deter migration; they do not prevent it if the customer concludes the relationship is exploitative, and the vendor's own migration deadlines have generated exactly that tension with parts of the installed base.

What does the artificial intelligence strategy actually consist of?

Embedding agents into existing business processes rather than selling models. The commercial logic is that a customer will not pay much for a general-purpose model and will pay for an agent that executes replenishment, forecasting or reconciliation correctly inside their own data.

The supporting acquisitions follow that logic. Recent purchases added an open data lakehouse platform for accessing non-SAP data without copying it, master data management for external sources, and tabular foundation model capability suited to structured business data rather than text.

The strategic claim is a data advantage: agents require governed access to business data, and the vendor holding the transactional system is better positioned to provide it than a model provider without process context.

The honest caveat, expressed by the company's own finance chief, is that artificial intelligence does not automatically resolve legacy data silos. Customers with fragmented, poorly governed data will not obtain good agent outcomes regardless of the model, which makes data readiness the actual bottleneck.

What do the 2026 numbers show?

Strong demand and compressed profitability. Second quarter results showed current cloud backlog of around twenty-three billion euros, up roughly twenty-six per cent at constant currencies, with cloud revenue growing in the low twenties and total revenue growing around a tenth.

Operating profit grew considerably more slowly, and full-year operating profit guidance was reduced by roughly a hundred million euros to reflect dilution from the recent acquisitions. Cloud gross margin declined slightly, attributed to one-off investment in test environments, agent capability and sovereign infrastructure.

The important signal was backlog growing faster than revenue, which had not occurred for several quarters and indicates that new contracted commitments are accelerating rather than decelerating.

For an analyst the question is whether the margin compression is investment or structural. Model inference costs are a genuine new variable cost in a business that previously had almost none, and outcome-based pricing has yet to demonstrate that it recovers them.

Why has Europe produced only one software company of this scale?

Because the company was founded when enterprise software was a systems and process business rather than a venture-funded consumer scaling business, and Europe was competitive at the former.

Enterprise software rewards domain knowledge, long sales cycles, reliability and proximity to industrial customers, all of which German engineering culture supplies. Consumer platform businesses reward rapid capital deployment, network effects and tolerance for years of losses, which European capital markets have not supplied, as the startup ecosystem pillar examines.

The implication is that Europe's realistic technology strength lies in business-to-business software embedded in industrial processes rather than in consumer platforms, and policy that targets the latter is likely to disappoint.

The secondary observation is that a single dominant company does not create an ecosystem by itself. It does supply experienced executives, and a meaningful share of German enterprise software founders have passed through it.

What should a CFO evaluating enterprise software take from this?

That the vendor's commercial model will change during your contract life, and that your leverage is highest before you commit. Subscription conversion, migration deadlines and consumption-based artificial intelligence pricing all shift economics toward the vendor over time.

The practical protections are contractual: price protection on renewal, defined limits on consumption-based charges, portability of data in usable form, and explicit terms covering what happens if a module is deprecated.

The second point is architectural. Customers who kept customisation minimal and used standard processes have migrated far more cheaply than those who built extensively on the platform, which reverses the conventional assumption that customisation creates value.

The third is timing. Migration costs rise as deadlines approach because consulting capacity becomes scarce, so the cheapest migration is the early one, and the most expensive is the one executed under vendor pressure in the final year.

How does the partner and consulting ecosystem work?

As a multiplier and a constraint. Implementation, customisation and support are delivered largely by system integrators and consultancies rather than by the vendor, which lets the vendor scale without carrying the services headcount.

The multiplier effect is powerful. Thousands of consultants trained on the platform create a labour market that makes the software safer to adopt, because customers can hire people who know it, and expensive to leave, because the skills are platform-specific.

The constraint is that the ecosystem has its own incentives. Integrators earn more from complex, customised implementations than from standard configurations, which historically pushed customers toward the heavy customisation that later made migration expensive.

The vendor has responded by promoting a clean core approach, keeping customisation outside the standard system in extension layers. That is technically correct and commercially awkward, since it reduces the consulting revenue of the partners the vendor depends on for distribution.

What does outcome-based pricing actually mean?

Charging for the result an agent produces rather than for the software licence or the compute consumed. A replenishment agent might be priced against inventory reduction achieved rather than against seats or tokens.

The appeal to customers is obvious: it aligns cost with value and removes the risk of paying for capability that does not deliver. The appeal to the vendor is that it captures more value where the software works well and creates differentiation against per-seat competitors.

The measurement problem is severe. Attribution of a business outcome to a single software agent is contestable, baselines can be disputed, and the customer's own execution affects the result, which makes contract negotiation and revenue recognition considerably more complex.

The likely commercial reality is hybrid pricing: a platform fee for access plus a variable component tied to usage or defined outcomes. Pure outcome pricing is rare in enterprise software for exactly these reasons.

What is the competitive threat over the next decade?

Not displacement of the core system, which remains extremely difficult, but erosion at the edges. Specialist applications for procurement, human resources, planning and analytics can be adopted alongside the core, and each one that succeeds reduces the surface area on which the incumbent monetises.

The artificial intelligence layer sharpens this. If agents can operate across systems through interfaces rather than requiring native integration, the value of holding the transactional system falls, because a competitor's agent can read and write to it.

The defensive response is to make the native agents genuinely better by using process context and permissions that an external agent cannot access, which is the strategy the company is pursuing. Whether that advantage holds depends on how open enterprise integration standards become.

What does sovereignty mean for an enterprise software vendor?

A product requirement rather than a policy position. European public sector and regulated customers increasingly require that their systems run on infrastructure under European control, which for a software vendor means offering a sovereign deployment option alongside the standard one.

That carries real cost. Sovereign environments have their own certification, staffing and release cycles, and the investment appears as margin pressure without corresponding revenue, which is one of the items management has identified as weighing on cloud gross margin.

The strategic return is access to workloads that would otherwise be unavailable, and differentiation against competitors who cannot offer it, a dynamic examined further in the digital sovereignty analysis.

Frequently Asked Questions