For much of the artificial intelligence boom, attention has focused on the companies building the models. The assumption was straightforward: as AI adoption accelerated, the biggest winners would be the new generation of model developers and AI-native startups.

Recent results from some of Europe’s largest technology companies suggest a more complex picture.

Established software, consulting, and cloud infrastructure groups such as SAP, Capgemini, Sopra Steria, and OVHcloud are reporting stronger demand, faster growth, or improved outlooks as businesses move beyond AI experiments and begin integrating the technology into their day-to-day operations.

The reason is becoming increasingly clear: accessing powerful AI models is relatively easy. Making them work reliably inside a large organisation is much harder.

From choosing AI models to making AI work

Large enterprises are unlikely to build their AI strategies around a single model or provider.

Instead, organisations are increasingly expected to use multiple models depending on the task, required performance, cost, security considerations, data sensitivity, and regulatory obligations.

This changes where some of the most important work happens.

The challenge is increasingly less about selecting an AI model and more about connecting AI with existing software, enterprise data, operational processes, permissions, and governance systems.

As UBS recently put it, “AI applications are the battleground, and that is where most value will be created.”

This environment plays directly into the strengths of established European technology companies. Long before generative AI became a strategic priority, these businesses were already helping large organisations integrate complex systems, modernise enterprise software, manage data, and operate technology in highly regulated environments.

Enterprise AI has to work with what already exists

Few large organisations have the luxury of starting from scratch.

Their technology environments often include software developed over decades, customised business applications, fragmented databases, legacy infrastructure, complex integrations, and processes shaped by years of operational requirements.

AI needs to work within this reality.

An enterprise AI application may need access to live company information while respecting user permissions. It may need to interact with several internal systems, preserve audit trails, comply with industry regulations, and fit naturally into workflows employees already use.

That complexity is becoming one of the main constraints on enterprise AI adoption.

Boston Consulting Group has found that AI deployment is progressing faster than many companies’ ability to manage it, while more than 70% of investors have expressed concerns about whether organisations have the technical and operational capabilities required to succeed with AI.

As businesses move from experimentation toward production, implementation, integration, data architecture, security, and governance are becoming increasingly important parts of the AI value chain.

Europe’s enterprise technology companies are seeing the impact

SAP offers one indication of this shift.

Its cloud backlog increased 26% at constant currencies to €22.9 billion as businesses continued moving critical finance, procurement, supply chain, and HR systems to cloud platforms that can increasingly provide the foundation for enterprise AI.

SAP’s acquisitions of data specialist Dremio and AI company Prior Labs also highlight an important part of this transition: enterprise AI is only as useful as its ability to securely access, understand, and work with business data.

European consulting and technology services companies are benefiting as well.

Capgemini raised its annual growth target after bookings increased 9.2%, while Sopra Steria upgraded its outlook following organic growth of 5.3%.

Their opportunity sits largely in what happens after an organisation decides to adopt AI: connecting models with business systems, restructuring data flows, redesigning workflows, introducing governance, and ensuring that new capabilities can operate reliably in production.

This becomes particularly important in industries such as defence, aerospace, healthcare, financial services, and critical infrastructure, where AI systems must operate alongside specialist software and within tightly controlled processes.

Sovereignty is becoming part of the AI infrastructure decision

Another trend could strengthen the position of European technology providers: companies increasingly want greater control over where and how their AI systems operate.

For organisations handling sensitive information, choosing AI infrastructure is no longer purely a question of computing capacity or price. Data location, jurisdiction, security, regulatory compliance, and technological sovereignty are becoming important purchasing criteria.

This is especially visible in sectors such as defence, aerospace, government, and critical infrastructure.

Airbus, for example, has selected Scaleway, the cloud provider owned by French telecommunications group Iliad, for sensitive industrial and defence applications, alongside AI technologies developed with Mistral. Around 70 critical Airbus applications are expected to run on Scaleway by the end of 2028.

OVHcloud is seeing similar momentum. Its public cloud revenue increased 20.2% in its third quarter, providing an early indication that demand for European-controlled cloud and AI infrastructure is beginning to translate into commercial growth.

For European organisations concerned about exposure to extraterritorial legislation such as the U.S. Cloud Act, locally controlled infrastructure can become an important part of the AI architecture.

The next phase of AI is about implementation

The current shift highlights a broader change in the enterprise AI market.

The first phase of the generative AI boom was dominated by models: which company had the most capable system, the largest context window, the best benchmark performance, or the fastest inference.

For businesses, those questions still matter. But they are only one part of the equation.

The next phase is increasingly about implementation.

Companies need to determine which processes should actually use AI, which data those systems should access, how models should interact with existing software, where human oversight remains necessary, and how security, permissions, monitoring, and governance should work.

They also need to modernise parts of their existing technology environments without destabilising the systems their operations already depend on.

This is why established European technology companies may occupy a stronger position in the AI economy than initially expected.

Their advantage is not necessarily in building the most advanced foundation model. It is in understanding the complicated environments where those models ultimately have to work.

What this means for companies

For organisations developing their own AI strategies, the lesson is important: access to AI is becoming commoditised faster than the ability to implement it effectively.

A successful enterprise AI initiative therefore needs to start with the operational environment rather than the model alone.

That means understanding existing workflows, identifying the systems and data involved, evaluating security and regulatory constraints, deciding where AI creates meaningful value, and designing an architecture that can evolve as models and providers change.

A flexible approach is particularly important. The AI model considered best for a particular task today may not remain the best choice two years from now. Enterprise systems should therefore avoid unnecessary dependence on a single model whenever the business case allows it.

The long-term advantage may belong to organisations that build the right foundations: clean and accessible data, adaptable software architecture, clear governance, strong integrations, and workflows designed around how people actually work.

For Europe’s established technology companies, that creates a significant opportunity.

The AI boom may have started with the companies building the models. But as artificial intelligence moves deeper into real-world business operations, an increasing share of the value could go to the companies capable of making those models useful, secure, compliant, and reliable inside complex organisations.

Source