Por: Julian Meitanis.- Director of Corporate Sustainability Services of KMPG
External contributor INFORSE
The debate about AI and energy has largely been framed around a single question: how much electricity does AI consume (and will in the future)? It is the wrong question. Or at least, it is incomplete.
The right question is whether AI’s energy demand, which is real, large, and accelerating, can be the catalyst that develops the clean energy infrastructure, which the world has struggled to scale quickly enough.
Starting with scale. The global electricity demand of data centers grew by 17% in 2025, well outpacing global electricity demand growth of 3%. Electricity consumption from AI-focused data centers grew even faster, by 50% [1].
The capital expenditure of five large technology companies (Alphabet, Meta, Microsoft, Amazon Web Services, and Oracle) exceeded $400 billion in 2025 and is expected to jump by a further 75% to USD 715 billion in 2026, now larger than global investment in oil and natural gas production [2].
By some estimates, data center energy consumption could approach 1,050 TWh by 2026, which, if data centers were a country, would make them the fifth largest energy consumer in the world, between Japan and Russia [3].
These are not marginal numbers. They represent a structural shift in energy demand.
𝗕𝘂𝘁 𝗵𝗲𝗿𝗲 𝗶𝘀 𝘄𝗵𝗮𝘁 𝘁𝗵𝗲 𝗵𝗲𝗮𝗱𝗹𝗶𝗻𝗲 𝗳𝗶𝗴𝘂𝗿𝗲𝘀 𝗮𝗿𝗲 𝗺𝗶𝘀𝘀𝗶𝗻𝗴
The same investment surge is beginning to reshape energy supply in ways that would not have happened on the clean energy sector’s own timeline.
The tech sector accounted for around 40% of all corporate power purchase agreements (PPAs) for renewables signed in 2025, and is now a major source of momentum for the nuclear and next-gen geothermal industries [4][5].
In practice, this means that AI infrastructure, with its 24/7 baseload demand profile and investable balance sheets, is becoming an energy catalyst. Or as IEA’s executive director Fatih Birol put it plainly: “….while AI is still an energy taker, it is also becoming an energy maker — driving forward innovative solutions like next-generation nuclear reactors, flexible data centers and long-duration energy storage”[6]. This is not a public relations narrative. It is a structural shift, seen through capital flows.
𝗦𝗼 𝘄𝗵𝘆 𝗶𝘀𝗻’𝘁 𝘁𝗵𝗶𝘀 𝗮 𝘀𝘂𝗰𝗰𝗲𝘀𝘀 𝘀𝘁𝗼𝗿𝘆?
Because the conditions that would make AI a net accelerator of the energy transition do not exist yet. Three governance gaps are standing in the way.
𝟭. 𝗘𝗻𝗲𝗿𝗴𝘆 𝗱𝗶𝘀𝗰𝗹𝗼𝘀𝘂𝗿𝗲 𝘄𝗶𝘁𝗵𝗼𝘂𝘁 𝗮𝗰𝗰𝗼𝘂𝗻𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆
AI leading companies publish ambitious net-zero commitments (e.g. IBM, Microsoft, Google). Most are not legally binding, not standardized, and not independently verified. Most disclosures remain vague and lack enforceable commitments on environmental impact.
The EU AI Act (effective from August 2025) has established a partial foundation: models assessed as posing systemic risk are subject to obligations, including energy consumption reporting [8]. But this covers a narrow slice of deployments and does not address GHG Scope 2 attribution, the most consequential question of whether the electricity powering a data center is from clean sources. We need energy accounting standards for AI infrastructure that are as rigorous as financial accounting. We do not have them.
𝟮. 𝗚𝗿𝗶𝗱 𝗶𝗻𝗳𝗿𝗮𝘀𝘁𝗿𝘂𝗰𝘁𝘂𝗿𝗲 𝗶𝘀 (𝘀𝘁𝗶𝗹𝗹) 𝘁𝗵𝗲 𝗯𝗼𝘁𝘁𝗹𝗲𝗻𝗲𝗰𝗸
The clean energy paradox is still: the renewable capacity exists, or can be built. The limiting factor so far has been building the grid infrastructure fast and flexible enough to deliver it. Planning and regulatory systems are being stretched by a wave of project applications for data centers, amid a broader trend of rapid load growth and electrification [9]. The result is that permitting queues and grid connection backlogs are becoming the de facto energy policy for AI infrastructure.
𝟯. 𝗧𝗵𝗲 𝗴𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 𝗼𝗳 𝗔𝗜 𝗮𝗽𝗽𝗹𝗶𝗲𝗱 𝘁𝗼 𝗲𝗻𝗲𝗿𝗴𝘆 𝗶𝘀 𝗲𝘃𝗲𝗻 𝗳𝘂𝗿𝘁𝗵𝗲𝗿 𝗯𝗲𝗵𝗶𝗻𝗱
There is a second dimension to this problem that receives far less attention. AI is not just a large energy consumer; it is also being deployed across energy systems to optimize grids, forecast demand, manage renewable intermittency, and accelerate decarbonization. AI is already being applied operationally across energy systems. The IEA’s Energy and AI report finds that AI-based fault detection alone can reduce grid outage durations by 30–50%, and that widespread adoption of AI in power plant operations could deliver up to $110 billion in annual cost savings by 2035 through avoided fuel costs and improved maintenance[10]. But the governance of AI in operational energy systems is nascent at best.
How do we verify the emissions reductions attributed to AI-optimized systems? Who stress-tests and verifies the models? What happens when an AI-driven grid management system fails (with potentially consequential repercussions)? These questions do not yet have regulatory answers in any major jurisdiction.
𝗪𝗵𝗮𝘁 𝗴𝗼𝘃𝗲𝗿𝗻𝗮𝗻𝗰𝗲 𝗹𝗼𝗼𝗸𝘀 𝗹𝗶𝗸𝗲
The gap is not primarily technical; it is institutional. Three things need to happen, and none of them require waiting for perfect regulatory frameworks.
𝗙𝗶𝗿𝘀𝘁, energy accountability for AI should be treated as a financial disclosure obligation, rather than a voluntary commitment. EU regulation on CSRD / European Sustainability Reporting Standards is already demanding greater transparency from large European companies. The logical extension is mandatory, standardized reporting of energy intensity for AI model deployments.
𝗦𝗲𝗰𝗼𝗻𝗱, AI companies and energy infrastructure developers need a formal co-investment and planning framework, not just bilateral PPAs. Coordinated capacity planning should be managed through a coherent platform based on government and industry dialogue.
𝗧𝗵𝗶𝗿𝗱, the advisory and consulting community needs to stop treating AI and energy transition as separate service lines. The client asking for a net-zero roadmap and the client asking for an AI infrastructure strategy are increasingly the same client, making the same capital allocation decision. The professionals who can sit at that intersection, combining energy systems knowledge with governance expertise and an understanding of how AI actually works, are the people who will shape what comes next.
𝗧𝗵𝗲 𝘄𝗶𝗻𝗱𝗼𝘄
The AI boom could accelerate deployment and innovation in the power sector if spending continues and government support is aligned [11]. History suggests that energy transitions do not self-organize around technology momentum alone; they require the right regulatory architecture, the right incentive structures, and the right institutional capacity.
We are still in the early stages of getting things right. The capital is there, and the technology is advancing. The unequivocal demand signal, AI’s relentless appetite for reliable, clean power, is unlike anything the energy transition has had before.
What is missing is governance that matches the pace of the investment.
That’s where the work lies.
──────────────────
I am passionate about the convergence of energy transition and emerging technology, and convinced that the governance frameworks needed to make AI a net positive for the planet are still being written.
𝘝𝘪𝘦𝘸𝘴 𝘦𝘹𝘱𝘳𝘦𝘴𝘴𝘦𝘥 𝘢𝘳𝘦 𝘮𝘺 𝘰𝘸𝘯.
𝗥𝗲𝗳𝗲𝗿𝗲𝗻𝗰𝗲𝘀
[1] IEA (2026), Key Questions on Energy and AI. Paris: International Energy Agency. Available at: https://www.iea.org/reports/key-questions-on-energy-and-ai
[2] IEA (2026), Key Questions on Energy and AI. The capital expenditure of five large technology companies (Amazon Web Services, Google, Meta, Microsoft, Equinix) surpassed $400 billion in 2025.
[3] Brookings Institution (2026), op. cit. Estimate based on IEA central projections and independent analysis.
[4] IEA (2026), Key Questions on Energy and AI. The tech sector accounted for ~40% of all corporate PPAs for renewables signed in 2025.
[5] IEA (2026), Key Questions on Energy and AI. SMR pipeline with data centre operators grew from 25 GW (end-2024) to 45 GW (April 2026).
[6]Fatih Birol, IEA Executive Director, quoted in: IEA News Release, ‘Data centre electricity use surged in 2025’, 16 April 2026. Available at: https://www.iea.org/news/data-centre-electricity-use-surged-in-2025-even-with-tightening-bottlenecks-driving-a-scramble-for-solutions
[7]Thomson Reuters Institute (2026), ‘New data reveals AI governance gap between policy and practice, creating ESG risks’, 23 February 2026. Citing Harvard Law School Forum on Corporate Governance data. Available at: https://www.thomsonreuters.com/en-us/posts/sustainability/ai-governance-gap-esg-risks/
[8]European AI Act (Regulation (EU) 2024/1689), Article 51 obligations for general-purpose AI models with systemic risk, applicable from 2 August 2025. See also: Chen, H-Y. (2026), ‘AI Governance and Regulation 2026: A Complete Guide to Global Frameworks’.
[9]IEA (2026), Key Questions on Energy and AI. Planning and regulatory systems described as ‘being stretched by the wave of project applications for data centres, amid a broader trend of rapid load growth and electrification.’
[10] IEA (2025), Energy and AI, Chapter: AI for Energy Optimisation and Innovation. Paris: International Energy Agency. Available at: https://www.iea.org/reports/energy-and-ai/ai-for-energy-optimisation-and-innovation
[11] IEA (2026), Key Questions on Energy and AI. Paris: International Energy Agency. https://www.iea.org/reports/key-questions-on-energy-and-ai2025
