At a glance
- AI’s huge requirements for energy, and to a lesser degree water, are raising questions about practicalities and sustainability considerations.
- Technological advances and innovations could help alleviate the problem.
- Our research has identified companies at the forefront of innovation.
As the rollout of artificial intelligence (AI) and its supporting data centres gathers speed, it is becoming clear that expansion depends on significant infrastructure buildout – and interest has grown in the potential opportunity offered by enabling infrastructure. This encompasses a host of sectors from utilities to industrials through energy, real estate, materials and more.
At the same time, questions have arisen about practicalities and sustainability: with the paradigm shift in electricity demand, how can AI’s appetite for energy be met without a surge in carbon emissions? How can data centres be cooled without draining the water table in areas that are often arid? Could either of these challenges curtail AI’s growth ambitions? The answers lie with innovative companies such as clean power developers, grid and thermal management leaders, and water technology providers.
Power and emissions challenges
AI’s growth and data centres’ huge appetite for power will not only put upward pressure on carbon emissions but also boost efficiencies and investments in clean technologies, including nuclear. The rollout is reshaping power demand dynamics and data centres have emerged as a major driver of electricity consumption. Indeed, our forecasts suggest that by 2030 they could account for more than 20% of total US power use.
Carbon emissions will likely increase on a temporary basis but the exact volumes from AI data centres are difficult to measure. They are dependent on:
- The rise in emissions from fossil fuel use associated with growth in AI data centres.
- Emissions reductions brought about by the efficiencies and innovations that AI brings to the energy system and the economy at large.
Using in-house power forecasts, and assuming a portion of power for data centres will come from clean sources, we estimate emissions attributed to date centre power demand will represent a single-digit percentage increase in US carbon emissions. To mitigate these increases, we expect to see:
- Emission reductions from efficiencies and innovations brought about by AI implementation by companies across the AI ecosystem.
- Hyperscalers increasing investment in clean power, as well as other emerging clean technologies such as carbon capture.
Figure 1: Power-hungry data centres are driving clean energy demand
Source: IEA, World Investment 2025
Equally importantly, innovation in data centre technologies is leading to more effective use of power. These advances include advanced cooling technologies, modular system design, and AI-enabled controls. We expect this area of innovation to drive incrementally meaningful power efficiencies.
Water becomes a bigger consideration for AI growth
Paradoxically, power availability and fiscal incentives have led to a high number of data centres being concentrated in water stressed regions. In the US, Bloomberg estimates that 58% of facilities are in areas of high or extreme water stress. This is fostering friction with local communities and elevating permitting risk.
There are trade-offs between power usage and water consumption. For instance, liquid cooling paired with evaporative cooling for heat rejection from the data centre can deliver superior energy efficiency. However, based on our internal projections, even in a conservative scenario for AI growth, it could increase water demand for cooling fivefold by 2030. Conversely, air-cooled chillers or dry cooling alternatives can also be paired with liquid cooling, reducing water consumption for cooling. However these solutions may use more power, particularly in warmer climates, leading to higher regional water use associated with fossil fuel power generation. For this reason, it is vital that water and energy are considered together, with water-efficient cooling technologies and solutions such as the use of recycled wastewater adopted in arid regions.
Investment implications
With AI’s build-out happening at a breakneck pace the practicalities loom into focus. What is clear is that the companies developing innovations to alleviate the bottlenecks associated with power and water constraints may offer investment opportunities.
Our research has identified companies at the forefront of solving these problems. For instance, Schneider Electric and Eaton are pioneering energy efficiency solutions, as is Trane Technologies. TSMC is a global leader in water recycling for chip production. And companies such as Vertiv, Ecolab, Veolia, Xylem and Tetra Tech are involved with data centre cooling as well as supporting wider water treatment and recycling infrastructure.
These companies and others like them will play a key role in helping the hyperscalers turn their ambitions into reality. Additionally, the strong demand for renewables and storage is benefiting integrated utilities as we see greater pipelines from data centres, while the wave of investments in nuclear technologies to power AI sites will benefit nuclear-exposed companies at the front of the value chain. These include uranium producers and industrials players manufacturing reactors and small modular reactors.
Interested in learning more?
Read the key takeaways from our research trip to San Francisco, including meetings with tech heavyweights NVIDIA and Broadcom. We also explore how the AI infrastructure boom is creating new opportunities and risks for bond markets. On the power side, read our article on opportunities for investors in nuclear power, and you can also download a copy of our recent Thematic Insight on Power hungry AI and implications for the energy transition.
