Europe’s AI ambitions are increasingly becoming an infrastructure problem. Building larger AI models and deploying enterprise AI requires more than GPUs: data centers need reliable power, networks need higher bandwidth and lower latency, and governments want greater control over where critical workloads and data are processed. Against that backdrop, the IOWN Global Forum will bring policymakers, telecom operators, data center companies and technology leaders together in Amsterdam on October 1 for FUTURES Amsterdam 2026 – Building Europe’s AI Infrastructure.
Europe’s AI Strategy Is Becoming an Infrastructure Race
The next phase of artificial intelligence adoption will be determined as much by physical infrastructure as by algorithms.
Europe is investing heavily in AI capabilities, but scaling those ambitions requires a difficult combination of computing capacity, electricity, connectivity, data sovereignty and resilient digital infrastructure.
Those issues will take center stage at FUTURES Amsterdam 2026, a public session of the IOWN Global Forum’s Midterm Member Meeting. The event will take place October 1 in Amsterdam, bringing together infrastructure providers, cloud specialists, telecom companies, equipment manufacturers and European policymakers.
The event is significant because the infrastructure conversation around AI is broadening.
Early AI infrastructure discussions were dominated by GPUs and hyperscale data centers. Today, the bottlenecks increasingly extend across the entire technology stack—from electricity generation and grid connections to optical networking, cooling, data movement and regional computing capacity.
Omdia estimates that cumulative global data center investment could approach $1.6 trillion by 2030, while leading technology companies are expected to deploy more than $600 billion in AI infrastructure capital expenditure in 2026 alone.
That scale of investment makes infrastructure policy a strategic issue rather than a purely technical one.
AI needs more than compute
The rapid expansion of AI workloads is putting pressure on infrastructure that was not originally designed for today’s computing patterns.
Training and inference workloads move enormous volumes of data between processors, storage systems and users. As AI applications spread from centralized cloud environments to factories, hospitals, offices and other edge locations, the network connecting those environments becomes increasingly important.
The IOWN Global Forum has been advocating this broader infrastructure model through its AI Computing Continuum, an initiative developed with the Open Compute Project that seeks to connect cloud and edge computing through high-bandwidth, low-latency networking.
That reflects an important change in how AI infrastructure is being designed.
Instead of treating the data center as an isolated computing destination, infrastructure providers increasingly need to consider the complete path between compute, data and the end application.
For AI workloads, latency and network capacity can directly affect application performance.
Omdia has warned that networking is becoming a critical constraint for AI cloud providers, arguing that enterprises need to assess connectivity, security and resilience alongside raw compute capacity.
Power is becoming a strategic AI constraint
Energy is another major part of the equation.
The International Energy Agency expects global electricity consumption from data centers to roughly double to around 945 TWh by 2030 in its base case. Electricity consumption from accelerated servers, which are primarily driven by AI adoption, is projected to grow particularly rapidly.
Europe is not immune to that pressure.
The IEA expects European Union electricity demand to increase by around 2% annually through 2030, with data center expansion among the factors supporting additional demand.
That creates a difficult policy equation.
AI infrastructure needs to be deployed quickly enough to support economic competitiveness, but data centers also need access to reliable electricity without undermining affordability, grid stability or emissions targets.
The result is a growing connection between AI policy and energy policy.
Sovereign AI moves from concept to infrastructure
The Amsterdam event will also focus on sovereign AI infrastructure, reflecting another defining characteristic of Europe’s approach to artificial intelligence.
Sovereignty is not simply about developing European AI models.
It can also mean maintaining control over data, computing resources, networks and critical digital services.
For enterprises and governments, that raises questions about where AI workloads run, which companies operate the infrastructure, how sensitive information crosses borders and how much dependence exists on foreign cloud and technology providers.
The panel session, “Building Sovereign AI Infrastructure: From Ambition to Implementation,” will examine these practical challenges.
The issue is becoming increasingly relevant for telecommunications companies.
Omdia reported earlier this year that telecom operators across Europe and other regions are increasing investment in cloud infrastructure, data centers, GPU-as-a-Service and AI networking as demand for AI compute intersects with sovereignty requirements.
That potentially gives European telecom operators a larger role in the AI economy.
Rather than remaining primarily connectivity providers, telecom companies can become infrastructure partners for AI workloads, providing data center capacity, edge computing, network services and locally hosted AI resources.
Photonics enters the AI infrastructure conversation
IOWN’s involvement also brings optical and photonic networking into the discussion.
The Forum’s core proposition is that photonics can help move large quantities of data with greater capacity and lower latency than conventional electronic approaches.
IOWN describes its broader technology direction around photonics-based networking and computing, with the organization highlighting potential gains in transmission capacity, latency and power consumption.
This matters because AI systems are becoming increasingly communication-intensive.
A modern AI data center is not simply a collection of processors. Thousands of accelerators need to exchange information rapidly, while distributed AI systems increasingly connect resources across data centers and geographic locations.
That creates opportunities for optical interconnects, coherent optics, photonic switching and other technologies designed to move data more efficiently.
IOWN has also been exploring applications such as remote GPU connectivity and green computing, indicating that the Forum is positioning photonics as part of the practical infrastructure layer supporting future AI workloads.
Amsterdam provides a useful European test case
The choice of Amsterdam is also relevant.
The Netherlands has one of Europe’s most established data center and connectivity ecosystems, while the country’s position as a major internet exchange and network hub makes it an important location for discussions around digital infrastructure.
The event will include Stijn Grove, managing director of the Dutch Data Center Association, who will address the strategic role of data center infrastructure in the Netherlands and the infrastructure requirements behind Europe’s AI ambitions.
That local perspective is important because AI infrastructure cannot be planned purely at a continental level.
Power availability, grid constraints, land, cooling requirements, fiber connectivity, regulation and community acceptance vary substantially between countries and even individual regions.
A European AI infrastructure strategy therefore needs to translate high-level sovereignty and competitiveness goals into deployable projects.
The industry is moving from AI experimentation to infrastructure build-out
The timing of the event reflects a broader transition in the AI market.
Omdia reported that global cloud infrastructure spending reached $110.9 billion in the fourth quarter of 2025, up 29% year over year. The research firm expects global cloud infrastructure services spending to grow another 27% in 2026 as enterprise AI moves from experimentation toward production deployment.
That shift increases pressure on every layer underneath AI applications.
More production AI means more inference. More inference means more compute. More compute means additional data center capacity, electricity, cooling and networking. And as workloads become more distributed, connectivity becomes part of the application’s performance architecture.
This is why infrastructure discussions are increasingly moving beyond “How many GPUs can we deploy?”
The more consequential question is whether a region can build the complete ecosystem required to operate AI at scale.
From infrastructure ambition to deployment
IOWN Global Forum President and Chairperson Dr. Katsuhiko Kawazoe is expected to argue that infrastructure decisions being made today will influence technological leadership and economic resilience.
That proposition is increasingly difficult to dispute.
AI competitiveness will depend not only on access to advanced models and chips, but also on whether countries can provide the power, networks, data centers and regulatory environment required to operate those systems economically.
For Europe, the challenge is particularly complex.
The region needs to accelerate AI deployment while balancing energy constraints, sustainability objectives, cybersecurity, data protection and digital sovereignty.
FUTURES Amsterdam therefore represents more than another AI industry event. It reflects a broader shift in the AI conversation—from developing increasingly capable models to building the physical and digital infrastructure needed to make those models economically useful at scale.
The winners of the next stage of the AI race may ultimately be determined not just by who builds the smartest models, but by who can connect, power and operate them most effectively.
Market Landscape
The AI infrastructure market is expanding across several interconnected layers:
- AI compute: GPUs, CPUs, accelerators and AI servers remain the foundation of large-scale AI deployment.
- Data centers: Capacity, cooling, power availability and geographic location are becoming strategic considerations.
- Networking: High-bandwidth, low-latency connectivity is increasingly essential for distributed AI workloads.
- Optical infrastructure: Photonics and optical interconnects are being positioned as technologies for moving AI-generated data more efficiently.
- Energy infrastructure: Rapid data center expansion is increasing pressure on electricity systems.
- Sovereign AI: Governments and enterprises increasingly want local control over sensitive AI workloads, data and infrastructure.
- Telecom AI infrastructure: Operators are expanding into data centers, GPU-as-a-Service, cloud and AI networking.
- Edge AI: Computing is moving closer to factories, healthcare facilities, transportation systems and other physical environments, increasing demand for high-performance connectivity.
The competitive landscape consequently extends well beyond hyperscalers such as AWS, Microsoft Azure and Google Cloud. Telecom operators, data center providers, optical networking companies, semiconductor vendors and infrastructure specialists increasingly have roles to play in the AI stack.
Top Insights
- IOWN Global Forum will host FUTURES Amsterdam on October 1, bringing European policymakers and infrastructure leaders together around AI deployment challenges.
- Europe’s AI infrastructure challenge now extends beyond GPUs to electricity, data centers, optical networking, latency, resilience and sovereignty.
- AI-driven data center growth is increasing electricity demand, making energy availability a strategic factor in regional AI competitiveness.
- Telecom operators and optical-networking providers are gaining a larger role as AI workloads become more distributed across cloud, data centers and edge environments.
- The emerging AI infrastructure model connects compute, networks, energy and sovereignty rather than treating the data center as an isolated resource.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
