Artificial intelligence is no longer a revolutionary investment theme. Across public markets, the first-order AI trade has been recognised and, in many cases, richly rewarded.
Semiconductor leaders, data centre power and cooling providers, and electrification equipment manufacturers have all been pulled into the AI capital-spending cycle.
From a sustainability perspective, however, attention has centred on AI’s environmental and social costs, from data centre energy and water use to supply chain pressures and broader societal concerns. But can AI improve grid resilience, system efficiency and the economics of decarbonisation.
AI may add to the grid load, water demand, and emissions, but its relevance goes beyond the data centre footprint. Can the same technology also operate across the broader energy system to reduce demand, shift flexible load, increase utilisation of existing infrastructure and improve how renewable power is forecast, stored and dispatched?
The load: Externalities and bottlenecks
In just a few years data centres have become one of the clearest centres of gravity for incremental power demand. This growth carries a rising physical footprint in both emissions and water use.
The International Energy Agency (IEA) projects emissions from powering data centres rising more than 50% by 2030. Water use associated with AI systems was estimated at 313–765 billion litres in 2025.
For investors, that burden means credibility matters: operators, utilities, and suppliers will need to show that rising load can be paired with cleaner power, disciplined water use, resilient siting and capital plans that support decarbonisation.
If power access, ratepayer impact, water use, siting and interconnection are becoming conditions of approval, AI infrastructure will have to behave less like an inflexible load and more like an integrated grid participant – shifting workloads, using storage, supporting flexible interconnection and improving visibility into local constraints.
The levers: Enabling transition opportunities
AI’s climate relevance depends not only on the electricity it consumes but on the efficiency, flexibility and clean-power integration it can enable across the rest of the system. We see system-level opportunities that companies may pursue, organized around four practical levers.
Reduce
AI could reduce the burden by turning plant data into real-time operating decisions – tuning motors, pumps and compressors; identifying quality issues earlier; and improving maintenance schedules. The highest-value use cases are likely to sit in steel, cement, chemicals, aluminium, pulp and paper, and motor-heavy manufacturing, where efficiency, uptime, emissions and unit cost are tightly linked.
AI may help reduce electricity demand in buildings through smarter, occupancy-responsive control of heating, ventilation and air conditioning (HVAC) systems, lighting, and other equipment. Building-management software could help forecast occupancy, weather, equipment performance, electricity prices and grid conditions, translating forecasts into automated reductions in energy use.
Shift
While efficiency reduces the size of the load, flexibility shifts when it lands. AI may help the system use energy at better times, move some of the load off stressed hours and squeeze more value from the existing grid. This is a different kind of value: making the grid more adaptive.
Virtual power plants (VPPs) are the clearest expression of this flexibility value. A VPP uses software to aggregate customer-owned assets – including batteries, electric vehicle (EV) chargers, smart thermostats, heat pumps, rooftop solar and flexible commercial or industrial loads – into a resource the grid can rely on.
Unlock
If shifting demand extracts more value from when energy is used, unlocking capacity extracts more value from the infrastructure already in place. Across the US, Europe, China and other major economies, new demand and clean supply are arriving faster than networks can expand. Moving more power through existing infrastructure is a transition opportunity in its own right.
Grid analytics, sensors and control systems are often the fastest and lowest-cost way to increase utilization of existing infrastructure – and the fastest way to defer or downsize the infrastructure that would otherwise have to be built.
Optimise
Unlocking capacity extracts existing value from clean supply of energy; optimizing it means enhancing the payoff by coordinating increasingly complex power systems – forecasting generation, dispatching resources, managing storage and matching supply with demand in real time.
That task is increasingly important in markets with high renewable penetration, from Europe’s wind-heavy systems to China’s rapidly expanding solar fleet.
AI does not fix the underlying architecture. It sits on top of it, helping turn better data into better decisions. AI-enhanced weather forecasting could improve renewable output forecasts, helping grid operators plan reserves, traders price power and storage systems optimise charging and discharging.
The result may be a smarter coordination layer that reduces curtailment, improves storage economics and lifts the value of megawatts that reach the grid.
Separating leaders from laggards
We see the beginning of AI-enabled transition value being monetized across the four levers. A company does not need all four to be a transition leader. What matters is that the exposure is real, measurable, and linked to a durable business model rather than a pilot.
Anton Gorodniuk is a sustainability investment strategist at Allspring Global Investments. The views expressed above should not be taken as investment advice.