An opportunity for Europe
And this is precisely where Europe’s opportunity might lie: not in building the largest AI at any cost, but in creating the most trustworthy, resilient and environmentally sound AI infrastructure.
This shifts the strategic debate. In future, board members will no longer be discussing just model quality, licence costs or GPU availability. They will have to discuss access to energy, site planning, sovereignty, AI governance, cybersecurity and the resilience of their AI infrastructure.
For computing power is becoming a geopolitical resource. Anyone who cannot provide it reliably, securely and responsibly will remain dependent within the AI economy.
Computing power is becoming a matter of location policy
The global AI race is often described as a software competition. That is an oversimplification. In reality, a new logic of location is currently emerging.
Where large AI models are trained and operated, the demand for electricity, water and space is rising. AI needs more than just data. AI needs industry. And industry needs infrastructure.
The crucial question is therefore no longer simply: Who has the best model? But rather: Who can operate AI sustainably?
AI agents, autonomous systems, intelligent automation and generative AI will further increase the demand for computing power. The more AI becomes embedded in the core processes of public administration, industry, insurance, energy supply and critical infrastructure, the more important the question becomes: Where is this AI actually running? Using what electricity? Using what water? Under what governance? And under whose jurisdiction?
Anyone who fails to answer these questions is outsourcing the foundations of their digital future.
When AI controls critical infrastructure, the infrastructure itself becomes critical
This becomes particularly clear in the case of critical infrastructure. As long as AI merely summarises texts or assists with research, its failure remains merely a nuisance. However, as soon as AI is integrated into grid control, control centres, maintenance, situational awareness, incident response, water management, energy supply or transport systems, the risk category changes.
Then AI is no longer just a tool. Then the infrastructure behind the AI itself becomes part of the critical infrastructure.
It is then no longer just a question of available computing power. It is about reliability, traceability, cybersecurity, energy supply, access control, data sovereignty, operating models and resilience in the event of a crisis. AI that supports critical processes must remain controllable even in the event of a disruption. It must be auditable and transparent. And it must not depend on infrastructure whose failure would further exacerbate the crisis.
Europe’s advantage could, of all things, be green AI
Europe will not win the global AI race by copying the US. Capital markets, platform power and access to chips are too unevenly distributed for that. Europe’s opportunity lies elsewhere: in trustworthy AI, green infrastructure, regulatory clarity, industrial depth and digital sovereignty.
Modern data centres do not have to be environmental problems. They can become part of a sustainable infrastructure strategy if they are purpose-built in locations where renewable energy is available, waste heat can be utilised and networks are strategically planned.
Green AI is therefore not a romantic ideal. It is a matter of location policy.
Green AI belongs in the supply chain
However, green AI is not just an issue for data centre operators. It is also becoming relevant for companies that use AI. This is because AI services form part of the digital supply chain.
Anyone who integrates AI agents, cloud services, autonomous systems or intelligent automation into their processes is not just procuring software. They are also incorporating computing power, energy, water consumption, hardware, operating models and security architectures. This makes AI infrastructure part of the value chain.
Under ESG criteria, it may therefore become relevant in future to consider what energy sources are used to power AI, how efficiently data centres operate, how transparently resource consumption is reported, and whether providers can furnish robust information on sustainability, operational security and data processing.
This does not only apply to the sustainability report. It affects procurement, IT strategy, risk management, compliance, cybersecurity and digital transformation. Anyone procuring AI today must be able to explain tomorrow what environmental, regulatory and operational risks are associated with it.
Digital sovereignty begins at the power socket
Digital sovereignty is often discussed as a data issue. That remains true. But in the age of AI, this perspective is no longer sufficient.
Sovereignty does not lie with those who merely own their own model. Sovereignty lies with those who control the conditions under which AI can be operated: data, models, platforms, computing power, energy, security, governance and operations.
This means that computing power itself becomes a factor of power. It is not just about chips. It is about control over AI infrastructure, energy supply, operating models and security architectures. Those without a reliable AI infrastructure remain dependent. Those who do not develop sustainable energy sources for it remain vulnerable. Those who fail to establish AI governance lose trust.
Europe must not turn sustainably produced AI into a narrative of sacrifice or a morally charged commodity, but rather into a story of sovereignty, responsibility and strategic location policy.
Conclusion: What companies should be doing now
Companies should clarify which AI applications are being integrated into critical processes, what infrastructure dependencies this creates, and what ESG impacts are associated with their use.
This involves five specific steps:
Prioritise AI use cases: Where does AI really add value?
Assess infrastructure dependencies: What cloud, model, chip, energy and operator dependencies arise?
Assess KRITIS relevance: Which AI systems support processes whose failure would be critical to business operations or public services?
Request ESG data: What information do providers supply regarding energy consumption, water consumption, location, efficiency and waste heat utilisation?
Establish governance: Who decides which AI is operated where, with what security requirements and what recovery plans?
