For much of the artificial intelligence investment cycle, attention has centered on the newest pieces of the technology stack: advanced semiconductors, accelerators, cloud platforms and the models running on top of them. Yet some of the most important constraints emerging around AI are decidedly physical.

Processors consume electricity. Electricity becomes heat. That heat has to be removed continuously and reliably. Higher computing density requires more sophisticated power distribution, thermal management, electrical equipment, controls and engineering, and as the scale of AI infrastructure grows, technologies that have existed for decades are being redesigned around demands that barely existed a few years ago.

The result is an unusual intersection between old and new technology. Chillers, pumps, heat exchangers, electrical systems and mechanical contractors now sit alongside liquid cooling, intelligent controls, advanced sensors, predictive analytics and digital design tools. For the HVAC industry, AI is not simply creating another end market. It is changing the technical requirements of the market itself.

The AI Buildout is a Physical Infrastructure Story

The computing requirements behind AI are significantly more energy intensive than those of the traditional digital economy. The International Energy Agency estimates that global data-center electricity consumption will roughly double from approximately 485 terawatt-hours in 2025 to 950 TWh by 2030, with electricity consumption from AI-focused data centers growing faster still, roughly tripling over that period.¹

That growth has consequences well beyond power generation. Nearly every watt consumed by computing equipment eventually becomes heat that must be removed from the facility, and as more computing power is concentrated into individual racks, cooling becomes a greater engineering challenge.

Traditional data centers were largely designed around air cooling. AI clusters built around more powerful GPUs operate at dramatically higher rack densities, putting pressure on conventional thermal architectures. Vertiv notes that newer AI environments are moving beyond 100 kilowatts per rack, and that new cooling infrastructure is being designed around combinations of direct-to-chip liquid cooling and conventional air cooling.²

The distinction matters because cooling is no longer simply a matter of installing more air-conditioning capacity. Higher-density computing changes how entire facilities are designed. Cooling distribution units, pumps, heat exchangers, chillers, controls, backup power and electrical infrastructure have to function as an integrated system. Existing facilities may require substantial retrofits, while new facilities can be designed around hybrid or liquid cooling from the beginning.

This is where the distinction between “old economy” and “new economy” begins to break down. A chiller is hardly a new invention. Neither is a pump, an electrical enclosure or a ventilation system. But the operating requirements placed on those technologies are changing: greater cooling capacity within smaller physical footprints, tighter temperature control, greater redundancy and the ability to respond to workloads that shift quickly. Power availability, water consumption, energy efficiency and speed of deployment have become central design considerations.

Equipment manufacturers are responding with higher-capacity chillers and liquid-cooling systems, and electrical and thermal systems are more often designed together rather than as separate parts of a facility. Trane Technologies and Eaton announced an integrated thermal-management and electrical reference design for AI data centers in August 2026, reflecting a broader movement toward treating the infrastructure surrounding compute as one coordinated system.³

That convergence expands the AI infrastructure discussion beyond a narrow collection of technology companies. The companies building the next generation of digital infrastructure may include familiar industrial businesses alongside newer specialists in thermal management, power systems, connectivity, automation and advanced cooling. Continued advances in computing, in other words, are creating demand for innovation several layers removed from the semiconductor itself.

AI is Not Only Creating Demand for HVAC. It is Changing HVAC Itself.

There is another side to the relationship. Artificial intelligence and digital technology are beginning to influence how heating and cooling systems operate across data centers, commercial buildings and other facilities.

HVAC has gradually evolved from largely mechanical systems into networks of equipment, sensors, software and controls. The U.S. Department of Energy describes building controls as the “intelligent nervous system” connecting HVAC and other building systems; modern controls can monitor conditions, coordinate equipment, identify faults and adjust building operations dynamically.⁴

The technology continues to advance. Predictive controls can incorporate weather conditions, occupancy patterns, electricity prices and equipment performance. Machine-learning models can help detect abnormalities before equipment fails. Connected systems allow technicians to monitor facilities remotely, and digital design tools can model thermal performance before equipment is installed. The effect is to add a sophisticated technology layer to a very mature industry.

In September 2026, Trane introduced a set of AI and digital building-management tools spanning cloud applications, connected controls and thermal management. One company’s products should not be treated as representative of an entire industry, but the development illustrates how traditional HVAC manufacturers are incorporating software and analytics more deeply into their equipment and service models.⁵

That creates a circular relationship: AI requires more advanced HVAC infrastructure, while AI and software can make HVAC infrastructure more intelligent.

Modernization Extends Far Beyond Data Centers

AI may be accelerating innovation at the highest-performance end of the HVAC market, but the modernization opportunity is considerably broader.

The United States already has an enormous installed base of commercial buildings and HVAC equipment. The Energy Information Administration’s 2018 Commercial Buildings Energy Consumption Survey estimated approximately 96.4 billion square feet of U.S. commercial building space, with roughly 87.5 billion square feet using cooling.⁶ Much of that infrastructure was built before today’s connected controls, sensors and energy-management systems became commonplace.

That creates a different type of technology cycle. Unlike a software application that can be deployed almost immediately, buildings and mechanical systems are upgraded over long periods. Equipment reaches the end of its useful life. Older systems are replaced. Controls are added. Buildings are renovated. Energy costs change the economics of efficiency investments.

Digitalization can make those replacement cycles more meaningful. Department of Energy research indicates that high-performance control sequences can deliver roughly 30% average annual HVAC energy savings in commercial buildings, with similar reductions in peak HVAC demand.⁴ Lawrence Berkeley National Laboratory has separately estimated that nearly 30% of commercial building energy use is wasted because of equipment faults and HVAC controls problems, which it associates with roughly $17 billion in potential savings.⁷

Those figures illustrate why HVAC modernization involves more than swapping one mechanical unit for another. New equipment may bring improved efficiency, but the broader opportunity includes sensors, software, automation, commissioning and ongoing service.

The Opportunity Reaches Across the Industrial Ecosystem

Viewing HVAC solely through the lens of air conditioners and furnaces can therefore miss a meaningful part of the current industry. Today’s HVAC ecosystem intersects with electrical infrastructure, construction, engineering, building automation, data-center equipment, energy management, distribution and specialized services.

That broader definition is particularly relevant to the AdvisorShares HVAC and Industrials ETF (HVAC). The fund’s investment universe includes HVAC manufacturers and service providers alongside distributors, software and technology companies, energy-management businesses and other industrial companies.⁸ The common thread is the infrastructure required to manage temperature, power and increasingly complex physical environments.

AI adds another dimension to that universe. Building a data center does not end when servers arrive. Facilities require power infrastructure, cooling equipment, wiring, connectors, installation, engineering, monitoring and maintenance. Many of those functions are provided by businesses whose histories long predate generative AI. Their technologies may be established. Their end markets are changing.