Cheap Power Becomes the Binding Constraint on Enterprise AI Expansion
Artificial intelligence expansion has run into a physical constraint that no amount of capital can bypass: the availability of cheap, reliable electricity. Analysts now estimate that up to 50% of planned data centre projects globally may never be built due to insufficient grid capacity, a figure that should concern any investor exposed to the AI infrastructure buildout.
Energy Costs Are No Longer a Rounding Error
A single ChatGPT query consumes roughly ten times the electricity of a standard search engine request. Multiply that across billions of daily interactions and the scale of demand becomes clear. Hyperscalers built their expansion models on the assumption that power would remain abundant and inexpensive indefinitely.
That assumption has collapsed. Brent crude has traded near 94 to 100 dollars per barrel amid renewed Middle East tensions and attacks on shipping in key maritime routes. Robert Staiger of the World Trade Organisation has warned that a sustained period of elevated energy prices could meaningfully slow the AI investment cycle. For operators, this means power procurement now sits alongside chip supply as a primary bottleneck.
Iberia's Strategic Position and the Nuclear Alternative
Spain and Portugal hold a genuine advantage here, though it remains underexploited. Pumped hydro storage capacity across the peninsula offers a buffer against renewable intermittency that few other European markets can match, yet investment in expanding this asset class has lagged behind solar and wind deployment.
Finland has moved decisively, with Fortum and Google signing a strategic agreement to expand nuclear-backed generation capacity specifically to serve new data centre load. This model, pairing firm low-carbon power with hyperscaler demand, offers a template that Spanish and Portuguese grid operators and policy advisors should study closely. Iberian nuclear capacity, alongside expanded pumped hydro and green hydrogen integration, could position the region as a competitive host for AI infrastructure if grid connection permitting is streamlined.
Under EU ETS rules, industrial electricity users face rising compliance costs as free allocations continue their scheduled reduction through the current trading phase. Data centre operators sourcing power from gas-backed grids will increasingly absorb carbon costs directly, sharpening the business case for renewable PPAs and nuclear offtake agreements over spot market exposure.
What Investors and Operators Should Track
The "show me the money" phase of AI investment means capital is now demanding measurable returns, and energy costs directly compress margins on inference and training workloads. Investors should monitor grid connection queue times across Spain and Portugal, EU ETS free-allocation phase-out schedules affecting industrial power users, and whether Iberian pumped hydro projects secure funding through national recovery and resilience mechanisms.
Operators planning data centre capacity in the region should prioritise firm power contracts over speculative renewable-only supply, given the volatility now embedded in gas and oil markets. The AI sector's next constraint will not be measured in GPUs but in gigawatts, and Iberia's energy mix gives it a narrow but real window to capture this demand before capacity elsewhere fills first.