AI, Energy, and the Climate Dilemma: The Transition from Burden to Catalyst

By Dr. Luke Soon

The Paradox We Face

Artificial Intelligence is simultaneously our most promising lever for sustainability and our fastest-growing source of energy demand.

The IEA projects that global data centre electricity consumption will more than double by 2030, reaching ~945 TWh – a figure equivalent to the current annual electricity use of Japan . In the United States, the EIA forecasts record-high demand in 2025–26, with AI data centres and crypto as key drivers .

And the mega-campus era is here:

xAI’s Colossus in Memphis, scaling towards hundreds of thousands – even one million GPUs . Microsoft/OpenAI’s “Stargate”, a proposed $100 billion U.S. supercomputer complex, signalling GW-class sustained load . Meta’s Manhattan-scale data centres, each demanding multiple gigawatts of firm, round-the-clock power .

Left unchecked, AI could account for 35–50% of total data-centre demand by 2030, overwhelming climate goals.

And yet — AI is also the key to solving the energy trilemma of reliability, affordability, and sustainability.

Where AI is Already Making a Difference

Supply Chain Visibility & Scope 3 Emissions AI can tackle the hardest problem in ESG: Scope 3 emissions, which account for 70–95% of corporate footprints . Tools like Altana AI deliver facility-level estimates, enabling compliance with mechanisms like the EU Carbon Border Adjustment . Circular Economy & Resource Efficiency SK Ecoplant’s AI-robotics project has cut EV battery dismantling time by 32%, with recycling efficiency gains projected at 30% by 2026 . Predictive maintenance extends the lifespan of industrial assets, turning waste into worth. Grid Optimisation & Renewables Integration The WEF/IEA note AI could save USD 110 billion per year in system costs by 2035 through better forecasting, predictive maintenance, and curtailment reduction . DeepMind’s work at Google cut cooling energy use by 40%, proving AI’s immediate potential in efficiency. Trade & Compliance The WEF’s AI for Efficiency, Sustainability and Inclusivity in TradeTech (2025) shows AI as a foundation for transparent, low-carbon supply chains, fraud detection, and sustainable trade finance .

The Net Zero Stalemate

We are at an inflection point. The Net Zero agenda, once the north star of corporate ESG, is now competing for attention with a trillion-dollar AI arms race. What was once greenwashing now risks morphing into AI-washing — with firms touting “AI for good” without accounting for the real energy costs of scaling models and data centres the size of Manhattan.

The uncomfortable truth: we will not hit climate goals by 2030 or 2050 unless we rethink the playbook for both AI and energy together.

Both the World Economic Forum (2025) and the Shanghai Institute of Technology white paper (2025) converge on one insight: AI can either accelerate or derail sustainability, depending on how it is deployed.

Optimising resource use: AI-driven precision agriculture can cut fertiliser and water waste by double-digit percentages, while predictive maintenance in manufacturing reduces downtime and extends equipment life . Reducing carbon emissions: Smart energy management systems have already lowered data-centre cooling loads by up to 40%, and city-wide AI traffic systems (e.g., Singapore) have cut congestion and emissions by ~20% . Promoting circular economies: AI-powered recycling (BASF, BlueOak, SK Ecoplant) boosts recovery rates, while AI-driven design enables more sustainable products (Adidas’ recycled-material shoes are a case in point) . Transforming trade: AI platforms optimise global supply chains, reduce waste, and enable green sourcing — Amazon’s logistics optimisation has cut costs 20%, while Maersk’s AI-driven route planning reduced vessel emissions by 12% .

But the same paper warns: without governance, AI adoption could deepen inequality, regulatory gaps, and algorithmic bias. Net Zero could fragment into isolated “islands of green” instead of a globally convergent pathway.

From Climate Burden to Climate Catalyst: The Transition Blueprint

The path to energy abundance must be deliberate, not accidental. Here are six imperatives, combining insights from the IEA, WEF, and Liu et al. (2025):

1. Governments must rewire planning for abundance

Mandate long-term transmission plans (e.g., FERC 1920 in the U.S.) and emulate China’s UHV grid strategy that co-locates AI compute with renewable-rich regions. Create regulatory sandboxes (TradeTech model) for AI in energy, trade, and carbon accounting .

2. Build firm, clean power alongside renewables

Extend the life of existing nuclear, accelerate SMRs/advanced reactors, and invest in geothermal. Don’t wait for AGI to discover “new physics.” Today’s tech stack is enough — what’s missing is political will and financing.

3. Require AI to optimise itself

Adopt energy-per-token KPIs before training runs. AI-driven carbon-aware scheduling (train in low-carbon hours/regions). Mandatory PUE ≤1.3 and water reuse in all new data centres .

4. Make circularity the default

Incentivise AI-enabled recycling, reuse, and circular product design. Adopt digital IDs to verify sustainable inputs across global trade flows .

5. Measure what matters

Shift from annual offsets to 24/7 Carbon-Free Energy (CFE) with granular certificates. Require Scope 2 (hourly) and Scope 3 disclosures across supply chains, enabled by AI mapping .

6. Embed equity and inclusion

Subsidise SMEs’ access to AI (to avoid a two-tier economy). Deploy AI in underserved communities (healthcare, education, mobile finance) — as Liu et al. highlight, this is how AI’s dividends spread beyond elite firms .

Why Current Climate Goals Are at Risk

Even with record renewable deployments, the math doesn’t balance.

AI loads are:

Lumpy (gigawatt-scale per campus) 24/7 (not just peaky or seasonal) Rising faster than transmission build-outs can catch up

Without firm clean power, AI growth risks higher marginal emissions, curtailment of renewables, and local community backlash over land and water use.

The Transition Blueprint: From Burden to Solution

Here’s the 12-point playbook I believe policymakers, boards, and hyperscalers must adopt:

24/7 Carbon-Free Energy (CFE), not annual offsets. Hourly, locational matching should become the gold standard. Google’s 24/7 CFE work is the benchmark . Firm Clean Power in the Mix. Nuclear is unavoidable. The U.S. brought Vogtle Units 3 & 4 online in 2023–24 (2.2 GW). China has ~30 reactors under construction. Advanced reactors and SMRs should be tied to AI hubs . Long-Duration Storage (LDES). Beyond lithium-ion: CO₂ batteries (Energy Dome), flow batteries, and gravity storage to time-shift renewables for AI’s round-the-clock demand . Smart Siting. Follow China’s “East Data, West Compute” model: co-locate compute with stranded renewables and build UHV lines to demand centres . Transmission First. Back FERC’s Order 1920/1920-A to accelerate regional transmission in the U.S. Without wires, clean generation is stranded . Efficiency Gates for Scaling Models. Introduce “energy-per-token” metrics before greenlighting new model training runs. Sparsity, mixture-of-experts, and distillation must be defaults, not afterthoughts. Water and Heat as ESG Constraints. Mandate non-potable cooling and circular water loops. Incentivise waste-heat reuse for district heating. Circular Hardware Pipelines. Refurbish and recycle GPUs/servers. Use AI to manage end-of-life pathways and mineral recovery. Transparent Hourly Emissions Dashboards. Publish CFE scores per campus. Tie executive compensation to verified progress. Green Finance Innovation. Link loan covenants and SLBs to hourly carbon metrics, not just annual averages. Public-Private Sandboxes. Use zero-knowledge proofs and federated models for secure, privacy-preserving ESG reporting . National Strategies. U.S.: Strengths in capital, innovation, and nuclear restarts. Weaknesses in permitting and transmission. China: Coherence of national siting strategy, nuclear scale-up, and UHV grid gives it a decisive edge — albeit with water stress challenges .

The 12-step playbook to reach energy abundance (what to do now, not when AGI arrives)

1) Build a lot more clean, firm and flexible supply—fast

Scale renewables + storage on today’s tech curve (wind/solar/batteries), and stand up long-duration (thermal, flow, CO₂-compression, pumped hydro) with bankable offtake structures. Treat nuclear as a core pillar of firm zero-carbon: after decades of stagnation, the U.S. added Vogtle 3 & 4 in 2023–24 (2.2 GW). Costs and delays were real, but the plants will deliver 60–80 years of carbon-

Net Zero Reframed: From Target to Operating System

Net Zero is no longer a distant 2050 pledge. It must become an operating system for how AI is deployed, measured, and governed:

AI without clean power = emissions crisis AI without transparency = AI-washing AI without inclusivity = social backlash

When reframed this way, AI doesn’t undermine Net Zero — it becomes the engine to deliver it.

Voices Shaping the Debate

Peter Swartz (Altana AI) – pioneering AI-driven emissions tracking. Sebastian Klotz (PwC) – integrating AI into sustainable trade and customs. Rebecca Harding (Rebeccanomics) – framing AI for sustainable trade policy. Puay Guan Goh (NUS Business School) – thought leadership on AI and ESG ecosystems. Emmanuelle Ganne (WTO) – linking AI adoption to climate-aligned trade governance. Kate Crawford (AI Now Institute) – critic of AI’s hidden environmental costs. IEA & EIA – data on surging electricity demand from AI, forcing planners to rethink climate trajectories.

From Arms Race to Climate Race

Waiting for AGI to solve climate science is a dangerous illusion. The solutions we need — nuclear, long-duration storage, renewable scale-out, circularity, AI optimisation — already exist. The missing piece is alignment of incentives and enforcement of standards.

The choice is not whether AI is a burden or a blessing. The choice is whether we hardwire sustainability into the very DNA of AI’s growth.

If we succeed, the story of the 2020s won’t be about AI’s energy crisis. It will be about how AI was the catalyst that finally unlocked the era of energy abundance.

The AI arms race is real — but so too is the climate race. We cannot allow one to derail the other.

The truth is that AI is not the enemy of sustainability; unmanaged, it’s simply a mirror of our energy systems. If those systems remain fossil-heavy and inefficient, AI will accelerate failure. If we rewire them — through 24/7 CFE, firm clean power, efficiency gates, and transparent ESG — AI becomes the most powerful ally we have in solving the climate crisis.

The choice is stark but actionable: turn AI from climate burden to climate catalyst.