The following information was released by the
By multiple experts (3)
Chapter 1
Executive
Summary
The
Significant breakthroughs in generative AI in recent years have led to an acceleration of the focus on its potential as a general-purpose technology. Huge amounts of private capital have been poured into AI development, with Microsoft,
Data centres and supercomputers have become the factories of the future, storing information and providing the processing capabilities critical to the progress of AI. The magnitude of this is vast: 25,000
As the pace of AI accelerates and investment continues, energy consumption will rise. Globally, the
In the
Some of this demand will be difficult to predict, as the efficiency of AI will improve. But there is little doubt that the AI era will increase the need for power.
The significant energy demand of AI has raised concerns over its climate footprint and impact on the pace of the energy transition. While these are understandable concerns, there is another way of looking at the issue: countries and governments can harness the investment in and powers of AI to accelerate clean technology and the energy transition.
Alongside a growing demand for power there is a growing demand for this power to be clean. Many of the large AI hyperscalers also have ambitious climate strategies. Their pursuit of 24/7 clean-power matching and net-zero emissions has resulted in a boom in clean-energy investment.
The incentive to lead in AI is therefore creating an incentive to lead in clean energy. As laid out in a recent paper by the
As an example, companies such as
But the US need not be the only leader. The
However, the
For the
The
For the
Develop better-integrated strategic planning that brings AI considerations into the national Strategic Spatial Energy Plan and economic planning to more strategically identify sites for AI-infrastructure development.
Create a 21st-century energy grid by reforming the planning process, to ensure that decisions can be made at speed, and by regulating for rapid grid expansion and allowing more expansive use of innovative grid-enhancing technologies.
Create the conditions to attract investment into clean-technology projects by expanding power-purchase agreements, considering electricity-market reforms and ensuring that regulation permits new types of clean technologies to come on stream.
Encourage strategic siting and use of data centres by reforming the electricity market and considering the use of the planning system or the connections queue to prioritise certain projects.
Support RandD and startups to accelerate progress by expanding energy-efficient-AI RandD projects and working with academia and industry to expedite the translation of research into deployed solutions.
Chapter 2
Understanding
Data-Centre
Siting
and
Energy
Use
The data-centre business has evolved significantly over time due to advancements in technology and changes in business requirements.
In the early days of computing, data centres were primarily on-premises facilities owned and operated by large enterprises, supporting basic computing and storage needs, custom-built to enable specific organisational needs. The main priority for siting during this period was proximity to corporate headquarters to facilitate easy access and control. Most often, data centres were housed within the headquarters themselves.
The rise of the internet and the expansion of global business operations led to an increase in demand for data-centre capacity. This period saw the emergence of more standardised data-centre designs housed in custom-made buildings, which allowed businesses to scale their operations more easily. The infrastructure began to evolve with modular designs, increased focus on networking, and the introduction of server racks and mass storage systems. Power density started to increase as the need for more powerful computing resources grew. As the internet expanded, data centres began to move closer to network hubs and major cities to reduce latency and improve data-transmission speeds. Proximity to fibre-optic networks became critical, and there was a growing need for facilities in multiple geographic locations to support redundancy and disaster recovery.
The emergence of cloud computing dramatically transformed the data-centre business. Companies like Amazon,
In the
With the rapid expansion of AI, the data-centre industry is changing further. By 2028, more than half of data centres will be dedicated to AI, with their development being designed specifically for AI workloads. Leading tech companies have announced more than
AI-specific data centres will need to be capable of providing larger amounts of power and removing it when it turns into waste heat. Ten years ago, nearly all data centres used fewer than 10 megawatts (MW) of power. Today, large data centres can require 100 MW of power or more equivalent to powering more than 230,000 homes. While the industry continues to innovate on chip efficiency, AI-specific chips are consuming more absolute power. For instance, while
This growing demand for power is combined with an increasing focus on sustainability, with a push towards carbon neutrality. Innovations in energy storage, such as battery technology, and the integration of AI and machine learning for predictive maintenance and energy optimisation are becoming more common. Data centres are also exploring more efficient cooling methods and are increasingly relying on renewable energy sources. This means that regions with abundant renewable energy, like the Nordic countries, are becoming preferred locations.
Additionally, geopolitical factors, data-sovereignty laws, proximity to 5G networks, and local regulatory environments and incentives such as tax breaks are influencing where new data centres are being built. For instance,
As data-centre workloads become increasingly AI-specific, the future of data centres will follow a hybrid model. Edge computing facilities will be located near major cities to handle low-latency tasks, while hyperscale data centres, which power large-scale computing and AI, can be situated in more remote locations optimised for energy efficiency and cost. Colocation data centres, data-centre facilities that rent out rack space to third parties for their servers, will continue to serve as critical hubs within this hybrid model, bridging the gap between edge and hyperscale facilities. They will provide businesses with the flexibility to locate near urban centres for low-latency requirements and in more remote areas where energy costs are lower, ensuring a balanced approach to data management. Edge data centres will process data close to users, reducing network load and ensuring quick responses, while hyperscale facilities will manage extensive data storage, complex computations and AI training. This approach allows for both rapid, localised processing and the efficient handling of massive workloads, ensuring the seamless delivery of diverse digital services. Together, edge, hyperscale and colocation data centres will form a complementary network, meeting the growing demands of real-time applications and large-scale computing.
This more decentralised, hybrid approach means data centres could serve important functions beyond just that of delivering critical digital services. For instance, data centres at the edge could be integrated into new housing-development plans, providing lower cost energy and heating to local communities through waste-heat recovery efforts something a number of Nordic countries are already piloting. Hyperscale data centres could be more deeply integrated into the grid, becoming more demand responsive and active in helping balance the grid. In regions with easy renewable-energy access, data centres could be directly integrated with vertical-farming efforts, leveraging waste heat, sharing infrastructure and reducing operational costs.
Data centres can become more than just the sum of their parts, but doing so requires joined-up thinking across sectors, and smarter and more strategic planning and investment strategies.
Chapter 3
AI’s
Energy
Requirements
and
How
the
Stacks
Up
As energy access becomes an increasing bottleneck in AI-data-centre development, understanding the specific energy requirements of AI is crucial to identifying the appeal of a location for investment in AI data centres.
First, rapid access to abundant and uninterrupted power is essential. This means having the requisite grid infrastructure in place to allow for fast connections, as well as consistent access to power to limit interruptions in real-time operations and allow for continuous learning.
Second, energy should ideally be low in cost. Given the vast amount of energy that data centres consume and the fact that electricity costs constitute a significant portion of the operating expenses, AI-data-centre providers are cost sensitive. High electricity costs can erode profit margins, particularly for colocation data centres. This means that AI-data-centre providers are looking for low-cost locations.
Together, the need for secure and reliable low-cost power requires as much insulation as possible from geopolitical risks or weather disturbances that lead to price volatility or supply interruptions. The primary solution to this is strong access to domestic sources of power, and ideally baseload power sources such as nuclear or geothermal, or solar combined with batteries in areas with high solar outputs.
Finally, the demand for this power is increasingly shifting to clean sources. Many AI companies, particularly hyperscalers, have made commitments to dramatically reduce their carbon emissions. Despite differences in approaches,[_]
The countries that can provide rapid access to clean, stable, low-cost energy are the contenders to become real AI superpowers.
The
However, the system remains plagued by a historic lack of anticipatory investment in the electricity grid and long planning processes, leading to a long connections queue for both renewable-energy projects on one side and data centres on the other. The
With ambitious clean-energy and AI strategies, the
Chapter 4
The
Keys
to
Success
To become a leader in the interlinking fields of AI and energy requires the
Integrate Strategic Planning
As laid out in TBI’s Greening AI: A Policy Agenda for the Artificial Intelligence and Energy Revolutions, the starting point to becoming a leader in AI is better coordination and strategic planning on a national level.
As outlined above, the concentration of the
But with the development of more diverse data-centre workloads, there are significant opportunities to enable more strategic placement of AI data centres across the
For instance, encouraging more data centres to be built in areas where renewable power is plentiful and less grid infrastructure is required, such as
But enabling infrastructure with the right strategic planning driving it must be in place.
The
This plan should also include AI infrastructure in its remit. In practice, this would mean integrating decisions on energy supply, grid infrastructure and communications infrastructure, such as 5G.
For instance, it could identify sites for small modular reactors, for instance on old nuclear or coal sites, where data centres could buy energy from them, encouraging development of low-cost and reliable energy that meets developers’ requirements. It could also mean encouraging strategic siting of data centres in areas where waste heat from operations could power local heat networks.
Tools powered by data and AI, such as digital twins, can enable this. By integrating data about the physical world from several sources, a digital twin can be used to test and model different scenarios and understand complex whole-system challenges. The government could integrate information about grid, generation, communications infrastructure, population centres and land-use to identify strategic sites for development.
This requires data centres and AI-growth demands to be integrated not only into the SSEP, but also into other planning efforts across government. Data centres should be considered as a cross-functional, utility-like resource in any planning-development effort; they could even be considered the central point of potential special economic zones, with a focus on building cluster and wider agglomeration in other critical industries such as biotech. Joined-up thinking and planning across departments and ministries is essential to tap into the wider opportunities they offer. Connecting spatial energy plans directly to municipal planning efforts, including district heating and housing development planning, would be a critical first step in doing this.
Create a 21st-Century Electricity Grid
Once the strategic plans are in place, building out new and strengthening existing grid infrastructure needs to happen quickly and reliably.
Currently, it takes more than a decade to build a new transmission line in the
This is another place where the planning system needs reform.
TBI has previously set out necessary changes to the
More straightforward planning consent is also required for data centres themselves. The new government has already announced its intention to include data centres in the NSIP regime, which, combined with radical reform of the NSIP regime, would also ensure rapid development of data centres.
Alongside planning reform, wider reform is needed to rapidly build the grid capacity the country needs to connect renewable energy projects with electricity demand, like that of data centres. The regulatory system around the grid must be set up for the transmission network operator and distribution network operators to invest in expansion and optimisation of the grid and to be empowered to rapidly and effectively connect new projects.
There are also ways to make the most of the existing grid to meet immediate demand. These efforts, such as the adoption of new technologies like dynamic line ratings or high-performance conductors, need be prioritised and expedited.
To do this, Ofgem should be given a similar remit to
Ofgem and NESO should also accelerate progress on reforming the connections-queue system to better weed out projects that are delayed or will never come to fruition and make the connections queue more dynamic. To help drive accountability and allow for better analysis of the connections queue, better data on renewable-energy and data-centre projects in the connections queue should be made publicly available.
Create the Conditions for Investment
Leading the AI and clean energy nexus will require the
The hyperscalers have significant funding to put into clean energy and are already deploying it at scale.
Microsoft is also investing in fusion companies to accelerate the development of clean abundant sources of power. In March,
To do this, the
The government could also consider whether there are ways to drive investment into grid infrastructure as a part of these deals. For instance, while AI companies support the development of clean energy through the use of renewable-energy credits and offsets, questions remain as to whether these credits are driving new additional clean energy onto the grid.[_] The government should explore ways to incentivise investment in additionality through greener power-generation projects that would otherwise not exist both locally and beyond the domestic grid. This could be done, for example, through designing existing offset programmes that credit additionality investments.
To attract investment into clean-technology projects in the
Encourage Strategic Siting and Use of Data Centres
In addition to improved strategic planning and infrastructure delivery, the government should take proactive steps to encourage data centres to play a positive role in the energy system and make smarter siting decisions the easier and more attractive option.
For instance, to enable speed and clarity of decision-making, the government could use the SSEP to identify candidate zones of investment where simplified planning regulations would create clusters of investment. These zones could be combined with tax or grant incentives as outlined above.
The
The
Innovations are also happening in terms of how data centres use energy, with companies developing promising solutions that could support, rather than strain, the
The ESO and the government should take a proactive approach to assessing how innovative solutions could help augment the energy system and help create the conditions for arrangements to be developed to encourage these types of solutions. The
Finally, the government should better utilise the planning system and the connections queue to help prioritise projects that support the broader energy system. For instance, the connections queue could be used to prioritise projects that meet certain criteria as a means to encourage investment in and development of not only more energy-efficient technologies but also of approaches that bring additional renewable energy to the grid. Such criteria could include investments in renewable energy and grid additionality to help reinforce the grid locally,[_] utilising energy storage and energy management to integrate with the grid for energy export, integrated data-centre flexibility requirements in service-level agreements. Longer term data-centre projects with large-scale
Support RandD and Startups
Finally, new solutions are also necessary. Energy efficiency largely is and will continue to be driven by industry. While the amount of computing carried out in
But given the strategic importance of this issue, there are opportunities for the government to help accelerate progress and develop new avenues for efficiency.
Some efforts are already underway. Most notably, the
The government should review the case for expanding support for AI energy-efficiency projects. For instance, the
As set out in Reimagining the
Chapter 5
An
AI
and
Clean-Energy
Superpower?
The