AI Data Centre Energy Costs: What UK CIOs Must Factor In Now

AI infrastructure energy costs have outpaced technology budgets in 2026, making power demand the single most material constraint on UK enterprise AI strategy that CIOs did not model two years ago.

The pattern is now clear. Enterprise AI workloads — particularly those running large language model inference, real-time analytics, and GPU-accelerated processing — require rack densities and cooling infrastructure far beyond what traditional data centre contracts were written to support. According to research published by Intelligent CIO Europe, over half (52%) of UK IT leaders now cite rising power costs as their top data centre concern, ahead of uptime and vendor lock-in. That is a decisive signal: energy has become the binding constraint on AI scale.

Why Are AI Workloads Driving UK Energy Costs Beyond Forecast?

Traditional enterprise data centres consumed between 10 and 20 megawatts. AI-ready facilities now require 100 to 300 MW, and some hyperscale campuses are approaching 1 GW. The underlying driver is GPU density: average rack densities have risen from 8 kW to 11 kW within a twelve-month period across the UK’s enterprise estate, according to the same Intelligent CIO survey, and that trajectory has not plateaued.

The macro picture reinforces this. This figure is highly likely to be — a figure that the National Grid has flagged as material to both grid planning and carbon emissions projections. OpenAI’s decision in April 2026 to shelve a multi-billion pound UK data centre project, explicitly citing energy costs, confirmed what CIOs managing enterprise AI deployments already knew: power is no longer a commodity input. It is a strategic resource that must be planned, hedged, and governed with the same rigour as capital expenditure.

• Executive Action: Conduct an AI energy audit across all colocation contracts, on-premise infrastructure, and cloud commitments. Establish a baseline consumption figure for current AI workloads and project forward across your three-year technology roadmap.
• Identify which AI workloads can be scheduled for off-peak hours or consolidated onto lower-density infrastructure without compromising latency requirements.
• Brief the CFO on the total cost of AI ownership model, explicitly separating compute, storage, networking, and energy as discrete line items.

How Should CIOs Quantify the True Cost of AI Infrastructure?

Many UK CIOs built their AI business cases in 2023 and 2024 using cloud pricing assumptions that predated the current power and GPU scarcity environment. Those assumptions are now materially incorrect. The total cost of ownership of AI infrastructure must incorporate four elements that were often omitted from early investment cases: power purchase agreements or tariff exposure, cooling infrastructure capital cost, carbon reporting obligations under the UK’s Streamlined Energy and Carbon Reporting (SECR) framework, and the opportunity cost of constrained capacity.

For CIOs running hybrid or on-premise AI stacks, the financial exposure is direct. For those operating primarily through hyperscale cloud providers, the exposure is indirect but growing — cloud providers are passing energy cost increases through their pricing tiers, particularly for GPU-optimised instance classes. A revised total cost of AI model, reviewed quarterly and presented to the board’s technology or audit committee, is no longer optional governance. Under the UK Corporate Governance Code 2024, material technology risks — including those arising from infrastructure cost volatility — require board-level visibility and response.

• Executive Action: Remodel your AI infrastructure investment cases to include energy cost sensitivity ranges based on current National Grid pricing bands and projected AI Growth Zone tariff structures.
• Work with your CISO and sustainability lead to align AI energy consumption data with SECR and TCFD disclosures, as these are now expected to reflect data centre emissions specifically.
• Establish a cost-per-inference or cost-per-model-run metric for your top five AI workloads as a baseline for ongoing optimisation.

What Does the UK Government’s AI Growth Zone Policy Mean for CIOs?

The UK government has announced AI Growth Zones — designated areas where data centres will receive discounted electricity rates linked to excess wind generation capacity. The scheme is planned to come into effect from April 2027, meaning CIOs must begin planning now if they intend to site or migrate AI infrastructure to benefit from lower tariffs.

The policy reflects a broader government recognition that AI infrastructure is a national competitiveness issue. However, the April 2027 implementation date means enterprises planning new data centre investments in 2026 face a window of decision-making uncertainty: commit to existing colocation arrangements now, or defer investment to benefit from the new tariff regime. For CIOs advising boards on infrastructure strategy, this requires a clear position paper on timeline, site selection criteria, and the financial delta between current and projected post-AIGZ energy pricing. Review the government’s AIGZ consultation documentation and engage with your colocation providers on their zone eligibility status. Explore the INFORMD AI governance assessment to evaluate your organisation’s readiness to scale AI infrastructure responsibly.

• Executive Action: Map your current and planned AI infrastructure against the designated AI Growth Zone locations; assess whether planned capex should be deferred pending April 2027 tariff clarity.
• Engage your colocation and cloud providers for formal written positions on their AIGZ participation and expected energy tariff trajectories.
• Include AIGZ policy risk as a named item in your next board technology risk register update.

How Should CIOs Brief the Board on AI Energy Risk?

The board’s responsibility is not to manage power procurement — that remains a CIO and CFO operational matter. The board’s responsibility is to understand whether the organisation’s AI investment case remains sound under realistic energy cost scenarios, and whether the organisation’s energy consumption profile is being disclosed accurately to markets and regulators.

A board briefing on AI energy risk should cover three areas: the financial sensitivity of the AI infrastructure programme to energy price movements; the regulatory exposure under SECR and TCFD if data centre emissions are underreported; and the strategic optionality created or constrained by site and vendor decisions being made now. NEDs with audit, risk, or sustainability committee responsibilities should specifically request that AI infrastructure energy costs appear as a named line in the technology investment review, not buried in general IT operational expenditure. Use the INFORMD technology strategy review template to structure the briefing and ensure material risk is captured at the right level. Additional executive briefings on AI governance, infrastructure oversight, and regulatory compliance are available in the INFORMD briefing library.

• Executive Action: Prepare a one-page AI energy risk summary for the next board or risk committee, covering current energy spend on AI workloads, forward exposure, AIGZ policy optionality, and SECR/TCFD alignment.
• Request that the CFO include AI infrastructure energy costs in the organisation’s capital allocation framework review.
• Confirm with your legal and company secretary team that AI-related data centre emissions are captured correctly in the strategic report and directors’ report.

INFORMD provides intelligence briefings, tools and frameworks for senior business leaders across technology, finance, strategy and compliance. Based in Milton Keynes, UK, we help executives stay informed and act with confidence. Explore our full briefing library or access our free assessment tools.