AI growth to further strain data centre capacity

AI adoption is set to further increase data centre demand in 2026, pushing vacancy to record lows.  
Data centres underpin the digital world, serving as the fundamental physical infrastructure for AI services. The burgeoning demand driven by the growth of AI, combined with the already escalating digitisation of economies, will place further strain on the data centre industry. 

Vacancy rates within European data centres, across both primary and secondary markets, are projected to decline further this year. Having fallen below 10% for the first time in late 2024, the rate is forecast to reach an all time low of 6.5% by the close of 2026. This is due to surging demand, compounded by bottlenecks in the electrical grid, which limits the volume of new capacity that can be brought to market. Construction and overall project lead times are increasing, while short-term requirements from customers are also on the rise. 

Despite capacity constraints, the continued decrease in vacancy is occurring amid a period of record-breaking new supply entering the market. This year, we expect over 750MW of data centre capacity to be added across Europe. This volume is equivalent to the entire colocation data centre capacity of France as of 2025 being added in a single year. 

The combination of escalating demand, increasing but constrained supply, and grid limitations is expected to fuel continued upwards pressure on pricing and the need for innovative solutions. The industry is likely to see further strategic partnerships as operators strive to secure both power and capacity in a competitive landscape. 

Figure 9: Data centre vacancy rate among primary and secondary markets in Europe (%)

Source: CBRE Research
Note: 2025 value is estimate 

Trends to watch

  • On-site power

    The constraints of power availability, coupled with lengthy lead times for grid connections, are causing operators to explore on-site electricity generation as a practical alternative to reliance on the grid. Initially conceived as a transitional measure, some operators are now planning for long-term deployment. Despite certain disadvantages, such as more complex ability reporting and elevated costs when compared to the grid, we anticipate further proliferation of this technology, driven by growing tenant acceptance. 

  • Site selection

    Capacity at scale will become harder to source across Europe. The site selection process for tenants is becoming even more complex as viable alternatives are often spread across multiple submarkets, or even countries. Short-term availability commands a premium, while pricing is set to further increase due to a supply and demand imbalance. 

  • Tenant pool

    Alongside established hyperscale providers, a growing number of companies, such as GPUaaS providers, are now seeking double-digit megawatt capacity across Europe. This trend is expected to intensify, diversifying the wholesale tenant pool. These new entrants often have a limited operational history, potentially impacting their covenant strength. Consequently, robust risk management and evaluation strategies will become paramount for operators in the coming years. 

  • Short-term requirements

    Despite strong demand, most operators are unwilling to undertake speculative data centre builds given the significant CapEx commitment required. Partial pre-let agreements are now commonplace. Because of the anticipated rise in short-term capacity requirements from emerging companies, this mismatch is projected to exacerbate the supply-demand imbalance in 2026. 

  • Power density

    Power density requirements are escalating for specific applications, such as AI training or inference. The increase will further fragment the technological landscape within the industry, as some operators invest in infrastructure upgrades to accommodate these needs, while others focus on traditional low‑density infrastructure.