As European companies move artificial intelligence from demonstrations into operational systems, a growing share of the spending is reaching established software, consulting, and cloud suppliers rather than the businesses that built the underlying models.

SAP, Capgemini, Sopra Steria, and OVHcloud have all reported stronger demand, accelerating growth, or improved guidance during the latest reporting cycle. Their results indicate that access to an AI model is becoming one part of a much larger implementation market.

Large organisations still need to connect models to decades of software, fragmented databases, access controls, audit requirements, and business processes. That work is difficult to standardise, particularly in defence, healthcare, banking, aerospace, and public services, where an unreliable output or uncontrolled data flow can carry substantial consequences.

The change is creating a market for companies already responsible for enterprise resource planning systems, outsourced technology operations, systems integration, data platforms, and regulated cloud infrastructure. Europe has produced relatively few frontier-model developers, but it has a deeper base of businesses that manage complex technology estates.

Reported results show spending moving downstream

SAP’s current cloud backlog reached €22.9 billion in the second quarter, rising 26% at constant currencies. Cloud revenue increased 24% on the same basis, while revenue from its Cloud ERP Suite grew 27%.

The German software group is linking its AI strategy to business information held in finance, procurement, supply chain, human resources, and other core systems. Its acquisitions of data-platform company Dremio and machine-learning specialist Prior Labs also point towards the importance of preparing enterprise information for AI applications rather than treating model access as a complete product.

Capgemini increased its 2026 revenue-growth target after first-half revenue reached €12.08 billion and second-quarter bookings rose 9.2% to €6.55 billion. The consulting and services group has described a multi-year cycle in which organisations modernise applications, infrastructure, and data before deploying AI more widely.

Sopra Steria also raised its full-year organic growth forecast after second-quarter growth accelerated to 5.3%. Demand was particularly strong in defence, security, aerospace, and the public sector, where the French group has long-standing customer relationships and specialist operating knowledge.

At the infrastructure layer, OVHcloud reported that public-cloud revenue rose 20.2% on a like-for-like basis during its third quarter, reaching €65.6 million. Growth outside France more than doubled from the first half, while new availability-zone deployments in Paris and Milan supported corporate demand.

Together, those figures support a less theatrical account of the enterprise AI market. Companies are spending on cloud migrations, data engineering, software integration, workflow redesign, security controls, and governance because models cannot produce dependable business outcomes when disconnected from operational systems.

Sovereignty creates another source of demand

European providers are also benefiting from concern about where sensitive workloads run and which laws apply to the underlying infrastructure. Defence companies, public bodies, financial institutions, and operators of critical infrastructure increasingly need to account for jurisdiction, supplier dependence, and access to data alongside model quality.

Customers are not necessarily replacing US technology with entirely European stacks. Most large organisations are likely to use several models and cloud environments, choosing between them according to performance, cost, security, location, and regulatory requirements.

Instead, sovereignty is increasing demand for suppliers capable of integrating multiple systems while preserving a degree of customer control. Airbus, for example, has selected French provider Scaleway for sensitive industrial and defence applications and expects roughly 70 critical applications to run on its infrastructure by the end of 2028, while also working with Mistral on AI tools.

OVHcloud’s growth offers early commercial evidence for that preference, although sovereignty claims still require scrutiny. Hosting workloads with a European company can alter the jurisdictional position and reduce some dependencies, but it does not automatically make every hardware component, software layer, model, or subcontractor European.

The same caution applies to consulting demand. Integration work can generate substantial revenue while organisations build their first operational AI systems, yet the suppliers delivering that work are also using AI to automate software development, testing, documentation, support, and routine analysis.

Consequently, the implementation boom contains a potential margin problem. Established providers may win more AI projects while needing fewer people to complete lower-value tasks. Revenue growth will depend on whether they can move towards complex engineering, industry-specific systems, governance, and managed operations faster than automation reduces conventional billing.

Implementation becomes the competitive layer

The shift also alters the competitive position of model developers. A frontier model can offer strong performance across many tasks, but enterprise customers seldom buy models in isolation. They buy applications, redesigned processes, managed services, risk controls, and support arrangements that allow the technology to function within an existing organisation.

That creates space for software and services companies to remain between model providers and customers, selecting technology, controlling access, and embedding AI into systems they already manage. It also prevents the market from collapsing around one model vendor, because integrators can substitute different systems when cost, regulation, or performance changes.

The advantage held by Europe’s established suppliers is practical rather than glamorous. They understand legacy systems, procurement cycles, regulated operations, and the organisational compromises needed to put new technology into production.

That advantage may weaken as AI development tools reduce the cost of integration and model providers expand further into enterprise applications. Reported demand is nevertheless accumulating around the companies that connect AI to operational data and business processes, rather than only around those producing the models.