AI will use as much light as a hundred reactors in 2030: these will be the winners

The demand for artificial intelligence (AI) It does not grow, it devours. It does so just when electrical infrastructure is growing at an insufficient pace to keep up with the increase in consumption. Electrification is accelerating, air conditioning is exploding in half the world and emerging countries are building new factories. Added to all this is the largest expansion of data centers ever recorded. The result is beginning to be visible. More tension in the networks. More saturation. Higher prices. And more companies desperately looking for a way to get access to electricity before new chips.

but this movement is not explained only by a specific increase. According to data of the International Energy Agency (IEA), Global data center consumption could be between 620 and 1,000 terawatts hour in 2030. Translated, it is equivalent to the annual production of between 60 and 100 modern nuclear reactors.

This estimate has become a central reference for analysts and regulators, because it forces us to ask questionsyes that until recently remainedn in the background. Which sectors are prepared to accompany this jump in demand and which will become the new beneficiaries of a technological race that, unlike previous cycles, depends on complex physical infrastructures.

For years, attention has been focused on chip manufacturers and large digital service platforms. It stands to reason. They are the engine of growth. However, the massive deployment of AI models has shifted part of the focus to elements you haveAt the moment they barely appear in the presentations of results. lto electrical grid, industrial refrigeration and engineering able to put it all together.

Various reports of McKinsey and Bank of America Research describe the same scenario. The increase in consumption forces reinforce generation, transform substations, expand lines and replace equipment that have been operating for decades. The modernization of electrical and thermal infrastructure is becoming the critical point that will determine the rate of expansion of data centers during the coming years.

The first bottleneck

lThe story begins on the internet. The increase in connection requests from digital projects has exceeded initial forecasts in some territories. This has occurred in the United States, in regions where the concentration of data centers is high and available capacity is reducing faster than expected. The AI.E. argues that this pressure forces operators to prioritize more powerful transformersmedium voltage systems and more robust backup equipment.

This scenario opens a window of growth for electrical infrastructure manufacturers. bank of America Research mentions Eaton, Schneider Electric and GE Vernova as some of the groups best positioned to absorb the wave of orders because they cover essential components of the electrical flow. Covers everything from distribution boards to internal stability solutions.

Cool down the AI

The second front appears in temperature. Each new generation of chips consumes more energy and generates more heat. Rooms designed a decade ago no longer support that pace. McKinsey explains that liquid cooling is making strong progress in high-density projects because it guarantees constant temperatures and reduces the risk of interruptions. The difference is not minor. Maintaining stable servers that work 24 hours a day has become as crucial as having GPUs enough.

This transition is driving companies specialized in advanced industrial air conditioning. Vertiv, trane Technologies and Johnson Controls They appear recurrently in sector reports due to their ability to integrate complex thermal solutions. Demand comes from both new facilities and operators who wish to update centers already in operation to avoid falling behind in energy efficiency and operational continuity.

The industrial base that supports growth

The third block is found in specialized engineering and construction. Building a modern data center involves large-scale civil works, precise electrical installations and control systems that cannot fail. The supply of companies capable of coordinating this type of projects is limited and the order book is growing rapidly. The analyzes consulted highlight EMCOR group and Quanta Services for his track record in critical infrastructure integration and for its presence in projects where demand already exceeds installed capacity.

Along with them, manufacturers of industrial generators are experiencing a moment of special prominence. In a context in which the network takes years to expand its capacity, Operators seek to ensure energy redundancy to avoid any risk of interruption. Caterpillar and Cummins have reinforced their production due to the increase in orders for support systems that accompany new centers and expansions under development. The goal is simple. Ensure power is available even when the grid is not yet at peak performance.

ANDthis set of sectors It forms an ecosystem that until recently occupied a discreet place in the technology market. The situation has visibly changed. The IEA describes a transition where digital innovation and industrial modernization advance at the same pace and need each other. The demand forecast for 2030 forces us to coordinate electrical investments, thermal infrastructure and network expansions with a complexity that did not exist in previous cycles.

This is the point at which the two major trends of the moment come together. On the one hand, the expansion of the AI. On the other hand, the need to adapt the physical infrastructure to an electrical consumption which approaches the combined output of one hundred nuclear reactors.