The construction of new data centers and the integration of AI solutions into business are impossible without transforming the energy sector. In the US alone, the amount of electricity consumed by data centers will triple by the end of the decade compared to current figures. This means massive investments — for example, to commission several data centers with a total consumption of 50 gigawatts (GW), over $500 billion will be required just for the infrastructure!
In the AI development story, the energy sector also sets the pace. The issue lies in limited resources: to build a powerful data center, it is necessary to scale the existing energy ecosystem, which includes reliable energy sources, infrastructure, electrical equipment in data centers, and highly skilled personnel. And while the rate of data generation continues to grow (last year alone saw a 22.5% increase, or 27 zettabytes), it may take up to three years to provide the energy for a new data center. This is the weakest link in AI’s spread: its potential cannot be realized without access to the corresponding energy infrastructure.
According to McKinsey, generative AI (GenAI) will help the economy gain between $2.5 and $4.4 trillion in the short term. To achieve this, several data centers with a total capacity of 50 to 60 GW need to be built in the US alone. As a result, data center expenses are already rising — according to Gartner, they increased by 24% last year! This is the direct impact of GenAI and its growing demand for computational power.
The specific energy demands of data centers differ from industrial enterprises that also operate 24/7. For example, data centers have backup energy storage systems for continuous operation, and their owners already pay more for electricity than the average market rate.
The increase in demand driven by AI has already led to a shortage of computational power due to difficulties in connecting to new power grids. At the same time, the pace of electricity generation is rising in sync with the number of data centers being commissioned — so the problem lies with energy infrastructure, not with the quantity of energy itself. Another limiting factor is the difficult situation with staffing: there is a shortage of electricians and workers at manufacturing plants that produce semiconductors and batteries.
The International Energy Agency (IEA) predicts that global demand for electricity in data centers will at least double between 2022 and 2026, primarily due to AI applications.
The Role of Investors in Successfully Implementing AI Projects
The further expansion of the data center ecosystem depends not only on the energy sector but also on the willingness of large investors to address the listed problems. They can help meet the growing demand for computational power from AI systems.
Providing Access to Sufficient Electricity
Investments are needed in the transmission and distribution of electricity — not only within the main data center markets but also in non-traditional locations. These could be investments in the expansion of hyperscalers, who are increasingly building their facilities near cheap alternative energy sources, or in new data centers in emerging markets, where construction times are shorter. Investors are also looking for opportunities to build power plants completely isolated from the grid, increase equipment density in existing data centers, or implement unconventional solutions, such as using nuclear energy.
Nuclear energy could see growing interest from the computing sector in the near future: nuclear plants are reliable and can provide stable generation at relatively low cost. While nuclear projects are complex and expensive, the continuing growth in demand (partly driven by AI) may make them attractive for necessary investment.
Producing Sufficient Energy Equipment
Investments in new and developing technologies will be required to address the shortage of critical equipment. Small companies that supply generators and power supplies are particularly in need of funding — without investors, they won’t be able to scale up and shift to working with hyperscalers. The trend of increasing the density of existing data centers is pushing for the production of more powerful equipment for rack installation — meaning resources are needed to modify existing product lines. Shifting to the production of modular solutions will allow for faster deployment and scaling of existing data centers.
Preparing Qualified Personnel
The trend of building new data centers in remote areas, where there is a lack of qualified personnel, coincides with the growing shortage of electricians. This presents opportunities for both companies specializing in finding or training personnel and for investors. They can fund the relocation of part of the production capacities outside the main facility to reduce the need for specialists on-site or find new approaches to recruiting and training staff. This will clearly become a key success factor for players providing computing resources.
We may soon see successful examples of this integrated strategy: Blackstone has announced plans to invest £10 billion ($13.4 billion) in the construction of a large data center in the UK focused on AI workloads. Another £110 million will be spent on upgrading the necessary infrastructure and retraining personnel for the facility.
Conclusions
Generative AI has impacted the pace of scaling and commissioning computational power more than any other technology since the early 2000s. However, new data centers are not being built as quickly as the rapidly developing technology requires — the construction of energy infrastructure is moving very slowly. The current situation can be corrected through investments in the development of next-generation power supply and data storage systems (their volumes are also growing rapidly). These are the areas where large players are willing to pay more than the market average to meet the growing demand for computational resources from AI systems.
