At the World Economic Forum in Davos last month, BlackRock CEO Larry Fink noted that data centres “cannot rely solely on intermittent sources like wind and solar”. He wasn't wrong. However, fossil fuel advocates claiming his comments justify a fossil-powered AI boom are. The challenge facing the industry is not fundamentally technical; it is a policy failure.
Data centres are a vital part of the global digital economy. As AI adoption becomes central to national growth strategies, this infrastructure’s footprint will only expand. Given their immense and fast-growing energy appetites, if the power sources remain carbon-intensive, digital growth will be pitted directly against national net-zero mandates.
Typically, there are two main culprits in a data centre for the high energy requirements: compute infrastructure (CPUs, GPUs and storage) and the round-the-clock cooling systems required to prevent the equipment from heating and throttling performance.
If the technology exists and the economics are improving, why are so few projects incorporating BESS?
Fink’s analysis of the data centre industry’s sustainability problem is correct. In a subsequent part of his speech, he acknowledged that the industry needs “dispatchable power because these data centres cannot simply turn on and off” – meaning that data centres require a 24/7 consistent flow of energy. However, because solar and wind are inherently intermittent, a surface-level analysis suggests a permanent role for fossil fuels.
This view, however, ignores the rapid maturation of Battery Energy Storage Solutions (BESS). These utility-scale battery banks stabilise renewables by saving excess power for later use. Year on year, BESS costs are plummeting. Crucially, it is also often faster to deploy dispatchable solar-plus-storage projects than to build equivalent natural gas capacity – allowing operators to plug into the grid and realise profits sooner if they shift to BESS. Theoretically, by “firming” renewable generation with large-scale BESS, it is technically possible to have sufficient renewable energy to power large data centres.
If the technology exists and the economics are improving, why are so few projects incorporating BESS? The answer lies in misaligned incentive structures.
Data centre operators are hypersensitive to industrial energy tariffs, which can account for up to 60% of total operational costs. Consequently, they site facilities wherever grid capacity is cheapest and most available, regardless of carbon intensity. If corporate mandates require “green” credentials, they often rely on carbon trading certificates – a paper exercise that doesn't change the physical reality of the grid and the environment they are located in.
Even though battery prices have dropped by 93% since 2010, in many growing global data centre hubs BESS projects remain relatively more expensive than fossil sources. Additionally, BESS requires significant upfront capital expenditure, which makes these projects harder to finance for individual developers.
The oncoming digital infrastructure boom does not have to derail us from the path of reversing climate change.
This is where policy can and must bridge the gap. National policies involving subsidies, viability gap funding, and streamlined permitting can incentivise adoption. India provides a compelling example as a country where sustained policy support has driven BESS prices to as low as $0.023/kWh. By clustering BESS and renewable projects near emerging data centre hubs, governments can further make the green choice the most profitable option. Once BESS projects are cost competitive, their relative advantage in electrification speed over fossil sources will likely reinforce market preferences further.
Beyond costs, current industry sustainability benchmarks measure energy efficiency using metrics like Power Usage Efficiency (PUE), but ignore the source of energy. A highly efficient data centre powered by a carbon-intensive source is still a climate liability. Here again, a policy-induced incentive structure tweak can bring about meaningful change. For example, policymakers can look towards Singapore’s Green Data Centre Roadmap and Green Mark for Data Centres policies, which attach preferences in project approvals to renewable energy integration and net-zero pathways, while also demanding better PUEs. These policy measures have helped cement Singapore as the hub for some of the most sustainable and efficient data centres globally while remaining a key data centre market.
The oncoming digital infrastructure boom does not have to derail us from the path of reversing climate change. The technology to decouple AI growth from emissions exists and climate advocates need not view data centre developers as the villains. Instead, industry and advocacy groups must mobilise political support to power the sustainable AI challenge through meaningful tweaks to policy.
