By Christophe Defert and Mike D’Aurizio

Every day, wind turbines around the world stop spinning and solar PV plants are turned off – not because of a lack of wind or sun, but because the grid cannot absorb the energy these generators produce.

This ‘curtailment’ issue is an expensive problem for the renewables sector. According to a 2025 report, €7.2bn worth of clean energy was generated across seven European countries in 2024 that was never paid for, lost to grid capacity constraints[1]. To take just one example: in Germany, solar curtailment roughly doubled in 2024, then again in 2025[2]. And it’s not just a European issue: in 2024, the state of California alone lost 3.4m MWh of renewable electricity, 29% more than the year before[3].

After twenty-five years of investment in renewable generation and storage hardware, costs have fallen to the point where solar and wind are now the cheapest forms of new electricity in most markets. Electric vehicles are reaching mass-market price points. Battery storage is scaling rapidly (108 GW of new battery storage capacity was deployed worldwide in 2025, 40% more than in 2024[4]). The hardware era has, in many respects, succeeded.

But this has created a new problem: how does a grid designed for predictable, centralised power generation adapt to managing millions of distributed, intermittent sources in real time?

Clearly investment in grid infrastructure is urgently required (Goldman Sachs estimates that approximately $720bn in grid upgrades will be needed by 2030[5]). But this is only part of the answer. To solve this incredibly complex coordination challenge, we also need to make our infrastructure more intelligent.

The intelligence opportunity

Fortunately, AI-enabled software is beginning to address this gap in ways that were simply not possible even five years ago.

For instance, demand forecasting platforms can now predict electricity consumption patterns with a precision that fundamentally changes how utilities and energy traders operate. Grid optimisation tools can dynamically balance supply and demand across distributed energy resources like rooftop solar, battery storage, EV charging and industrial loads. AI-based predictive maintenance systems can identify faults in transformers and transmission infrastructure before they cause outages; the IEA estimates that this could reduce outage durations by 30–50%[6]. And remote sensors, autonomous drones, and AI-based management can unlock additional transmission capacity on existing lines, without a single new cable being laid. The same IEA report suggests that AI-based grid management tools could unlock up to 175 GW of additional transmission capacity – equivalent to about 70% of the entire installed wind capacity of the European Union.

This clearly represents a substantial market opportunity. The global AI-in-energy market is projected to grow from around $9bn in 2025 to nearly $59bn by 2030, a compound annual growth rate of almost 37%[7] .

Demand forecasting, the application that most directly addresses the curtailment issue, is the fastest-growing segment within this. And the companies best-placed to capture this opportunity share a common characteristic: they were built by founders who deeply understand the operational reality of energy markets, as well as the technology.

Take Amperon, a US-based company we backed at Series A in 2022. Its founder spent over a decade trading power at Tenaska, EDF, and E.ON before building an AI-powered electricity demand forecasting platform that delivers three times greater accuracy than the industry standard. That domain expertise is exactly why utilities and energy traders trust Amperon with consequential decisions, enabling the business to increase revenue by more than 4x between 2022 and 2025.

Battery storage is another area where the economic logic is clear. AI-driven analytics – such as those developed by Accure, a German company we backed at Series A – can enhance battery performance by more than 20% and extend operational life by 25% or more. That is a material improvement in the economics of storage projects that are often valued over 15–20-year horizons, and a meaningful reduction in the materials and manufacturing cost incurred when replacing battery assets prematurely.

This opportunity is so timely and compelling for investors because it reflects the convergence of two significant forces. The first is the sheer volume of data now being generated by energy infrastructure – via smart meters, grid sensors, EV charging networks, weather stations, satellite imagery. And the second is the exponential increase in computational capability that enables us to process and act on this data.

Where intelligence meets impact

It’s also important to remember that the environmental impact of solving (or failing to solve) the grid intelligence problem is enormous.

Curtailment is not just a financial inconvenience for renewable energy developers. Every megawatt-hour of wind or solar power that is wasted because the grid cannot absorb it is a megawatt-hour that must be replaced by something else – usually gas peakers. More accurate forecasting means grid operators can integrate higher shares of intermittent renewables with confidence, use gas peakers less frequently, and move faster towards a cleaner generation mix.

That’s why the IEA projects that the widespread adoption of existing AI applications in energy and industry could deliver 1,400 million tonnes of CO₂ reductions by 2035 – equivalent to roughly 5% of global energy-related emissions in that year, and four times more than the emissions generated by AI data centres themselves. A 2025 study led by Nicholas Stern and co-authors at the LSE[8] puts the potential higher still: between 3.2 and 5.4 gigatonnes of CO₂e annually by 2035, driven largely by improvements in power systems, transport, and industrial efficiency. These are not purely speculative numbers, based on technologies yet to be invented: they are based on AI applications that exist today, applied to existing infrastructure. The constraint is deployment, not invention.

The pressure to solve the clean energy integration challenge is ramping up, particularly in Europe: the REPowerEU programme is targeting 42.5%-45%[9] renewable electricity penetration by 2030. By backing technology businesses that can make the grid more intelligent and efficient, investors have a huge opportunity to accelerate this transition and unlock economic value – while also helping to build a cleaner, more sustainable world.

This is the second in our series on the intelligence era of climate technology. Next: how AI and robotics are transforming sustainable manufacturing.

[1] https://beyondfossilfuels.org/wp-content/uploads/2025/05/REPORT_FINAL.pdf

[2] Federal Network Agency (BNetzA) via Clean Energy Wire: https://www.cleanenergywire.org/news/solar-power-curtailment-rise-germany-grid-expansion-lags-behind

[3] US Energy Information Administration: https://www.eia.gov/todayinenergy/detail.php?id=65364

[4] IEA: https://www.iea.org/reports/global-energy-review-2026/technology-battery-storage

[5] Goldman Sachs: https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030

[6] IEA, Energy and AI, April 2025

[7] MarketsandMarkets, 2025

[8] Stern, N. et al., Green and Intelligent: The Role of AI in the Climate Transition, npj Climate Action (Nature), June 2025

[9] https://energy.ec.europa.eu/strategy/repowereu-phase-out-russian-energy-imports/repowereu-4-years_en