The artificial intelligence boom is increasingly becoming a race to secure something far less visible than advanced processors: electricity, transformers and cooling capacity. While Nvidia and other chipmakers remain the most prominent beneficiaries of the surge in AI spending, the rapid construction of data centers is creating a parallel investment cycle in the industrial equipment needed to make those facilities operational.

That shift is changing the economics of the data center industry. Developers can purchase powerful processors and secure land, but those investments cannot generate revenue until enough electricity reaches the facility and the heat produced by dense computing equipment can be removed. As a result, suppliers of transformers, power-management systems, cooling equipment, generators and related infrastructure are becoming critical bottlenecks in the expansion of AI capacity.

The scale of the opportunity is enormous. McKinsey estimates that global data center investment could reach about $7 trillion by 2030, with AI-related infrastructure accounting for the largest share of the additional capacity required. Its estimates indicate that power infrastructure, cooling and other physical systems will absorb a substantial portion of that spending.

The investment opportunity is therefore spreading well beyond computing hardware. The companies positioned to benefit are increasingly those that can solve the physical constraints preventing more computing capacity from coming online.

Electricity Is Becoming The First AI Bottleneck

The most important reason power-equipment companies are benefiting is straightforward: AI data centers consume substantially more electricity than conventional facilities. As computing becomes more concentrated, operators require much greater power density within individual facilities, placing additional demands on transformers, electrical distribution equipment, backup generation and cooling systems.

The International Energy Agency estimates that global data center electricity consumption could more than double to around 945 terawatt-hours by 2030. Electricity consumption associated with accelerated servers, which are closely linked to AI workloads, is expected to grow particularly rapidly and account for a significant share of the additional demand.

This creates a mismatch between the speed of the technology industry and the speed of the electricity system. A new computing facility can potentially be designed and constructed within a few years, but transmission infrastructure, substations and grid connections often require much longer planning periods. The IEA has specifically warned that the broader energy system has longer lead times than data center development, creating a potential constraint on AI expansion.

That mismatch explains why transformers have become strategically important. A transformer is not an optional component that can be added after a data center is built. It is part of the electrical system required to convert and distribute power safely to the computing equipment. If a project cannot secure the necessary transformer capacity, construction progress elsewhere may have limited value.

The same problem applies to generators and other electrical equipment. Data center developers increasingly need multiple sources of power and backup systems because interruptions can be extremely costly for facilities running critical computing workloads. The result is a growing market for equipment that historically served utilities, industrial companies and conventional infrastructure projects.

Long Lead Times Are Turning Equipment Makers Into Gatekeepers

The AI infrastructure race is placing unusual pressure on manufacturers because demand is arriving faster than some traditional industrial supply chains can expand. Equipment manufacturers built around relatively predictable utility investment cycles now have to respond to hyperscalers that want capacity delivered much faster.

McKinsey estimates that data center infrastructure excluding information technology hardware could require more than $1.7 trillion in global spending through 2030. It also identifies power and thermal equipment as important areas where existing industrial suppliers may struggle to satisfy demand at the required speed.

This helps explain the large order backlogs reported by some power-equipment manufacturers. Companies with established production capacity, engineering expertise and relationships with large customers can secure orders well before facilities become operational. Their advantage is not necessarily that they invented a new AI technology. It is that they already manufacture equipment that the AI industry suddenly needs in much greater quantities.

The opportunity also extends beyond established manufacturers. Data center developers are looking for ways to reduce construction times, simplify installation and integrate power and cooling systems more closely. That creates room for companies capable of supplying complete systems rather than individual components.

The industry is therefore moving toward greater integration. Power equipment, cooling systems and computing hardware increasingly have to be designed together because the performance of one affects the requirements of the others. McKinsey notes that data center operators are placing greater importance on time to market and on suppliers capable of providing integrated equipment and services.

This could gradually alter the competitive structure of the industry. Companies that can deliver equipment quickly, integrate systems and provide maintenance may have an advantage over manufacturers competing primarily on the price of individual components.

Cooling Is Becoming As Important As Power

Electricity is only half of the infrastructure challenge. The more power that is converted into computing activity, the more heat has to be removed from the facility. AI accelerators operate at much higher power densities than conventional computing equipment, making traditional air-based cooling less suitable for some high-density deployments.

That is increasing interest in liquid cooling, which transfers heat more efficiently than conventional air systems in many high-density applications. McKinsey estimates that data center demand for cooling and other infrastructure will represent a meaningful portion of the overall growth in physical data center capacity.

The shift toward liquid cooling is important because it expands the number of companies participating in the AI infrastructure investment cycle. Manufacturers of pumps, heat exchangers, thermal management systems and specialised cooling equipment can benefit even if they have no direct connection to the production of AI processors.

Cooling also has an important economic dimension. Electricity used to remove heat adds to the operating cost of a data center. Improving thermal efficiency can therefore reduce both energy consumption and the amount of infrastructure required to support a given computing load.

This creates a feedback loop in the market. More powerful computing systems require more electricity. More electricity produces more heat. More heat requires better cooling. Better cooling requires additional equipment and engineering. As computing density rises, the physical infrastructure surrounding each processor becomes increasingly important to the economics of the entire facility. The consequence is that AI infrastructure is becoming less about simply installing more chips and more about designing an entire physical system capable of supporting them.

The Infrastructure Boom Has Its Own Risks

The growing opportunity for power and cooling companies does not mean every supplier will benefit equally. The surge in orders has already encouraged manufacturers to expand capacity, while investors are questioning whether some valuations have moved ahead of realistic long-term earnings.

Competition is another risk. High margins in attractive equipment markets encourage established companies to increase production and new competitors to enter. If data center construction eventually grows more slowly than expected, manufacturers that have expanded capacity aggressively could face excess supply.

Demand itself is also difficult to forecast. McKinsey’s data center investment estimates range from roughly $3.7 trillion to $7.9 trillion by 2030 depending on assumptions about AI adoption, efficiency improvements and capacity requirements.

That range demonstrates why the infrastructure opportunity should not be treated as a guaranteed multitrillion-dollar revenue stream for every company in the supply chain. Improvements in computing efficiency could reduce the amount of electricity required for a given AI workload. More efficient processors and software could also reduce the amount of physical infrastructure needed.

There is another constraint emerging in the United States: the difficulty of determining how much proposed data center demand is actually credible. Recent reporting has highlighted extremely large electricity requests from prospective projects, prompting some regulators to scrutinize whether all planned facilities are sufficiently funded or likely to be built.

That development matters because equipment manufacturers ultimately depend on real projects, not announced power requirements. A cancelled or delayed data center can push back orders for transformers, generators and cooling systems even when long-term AI demand remains strong.

The Winners Will Be The Companies That Solve Bottlenecks

The most important change created by the AI boom may therefore be the elevation of infrastructure from a supporting function to a strategic constraint. Nvidia can supply processors, cloud companies can finance data centers and developers can secure land, but none of those investments can produce computing capacity without reliable electricity and effective thermal management.

This is why power and cooling companies are increasingly being treated as part of the AI investment chain rather than as conventional industrial suppliers. The global build-out is creating demand for transformers, electrical distribution equipment, backup generation, liquid cooling, thermal management and increasingly integrated power systems.

The opportunity is especially strong for suppliers with proven technology, manufacturing capacity and established relationships with major data center operators. But the market is likely to become more selective as competition increases and project timelines become harder to predict.

The AI infrastructure race is consequently moving into its next stage. The decisive constraint may no longer be whether companies can obtain enough processors. It may be whether they can secure the electricity and cooling systems required to operate those processors at scale. That shift explains why a large share of the economic value created by the data center boom is moving into less visible parts of the industrial supply chain. The companies making transformers, cooling systems and power-management equipment are not merely benefiting from AI growth. They are increasingly helping determine how quickly that growth can physically happen.

(Adapted from TeadingView.com)

Categories: Economy & Finance, Entrepreneurship, Regulations & Legal, Strategy