The race to power artificial intelligence is creating a massive supply chain opportunity that extends far beyond chips and servers. VanEck has launched a new exchange-traded fund (ETF) designed to capture investment in the companies solving AI's most pressing infrastructure challenges: power generation, cooling systems, semiconductors, and electrical equipment.

Why Is AI Infrastructure Becoming Its Own Investment Category?

For months, the conversation around artificial intelligence has focused on software, applications, and which companies will dominate generative AI. But behind every large language model (LLM), every AI training run, and every inference serving billions of users sits an enormous physical infrastructure problem that few investors have fully appreciated. The five largest AI hyperscalers, companies like Microsoft, Google, Amazon, and Meta that operate massive data centers, are projected to spend approximately $750 billion on infrastructure this year alone. That figure dwarfs typical software spending and signals a fundamental shift in how the technology industry operates.

"Much of the AI conversation has centered on applications and software, but the scale of infrastructure required to support AI adoption is becoming increasingly important. What's happening now resembles a utility-scale industrial buildout more than a traditional software cycle, spanning semiconductors, energy systems, cooling technology and electrical equipment," said Nick Frasse, Product Manager with VanEck.

Nick Frasse, Product Manager, VanEck

The new VanEck Data Center Supply Chain ETF, trading under the ticker RACK, seeks to track companies across the entire data center supply chain. To be included in the fund, companies must generate at least 50 percent of their revenues from business segments tied to AI infrastructure and data center development. This represents a deliberate focus on the unglamorous but essential backbone of the AI revolution.

What Are the Actual Bottlenecks Holding Back AI Expansion?

The infrastructure constraints are real and measurable. U.S. data center power demand is growing rapidly, but the supply chain cannot keep pace. Power transformers, the electrical equipment that steps down voltage for data centers, now have delivery times stretching to as long as two to four years. This is not a minor inconvenience; it is a hard constraint on how quickly companies can build new data centers. If you order a transformer today, you may not receive it until 2028 or 2030.

McKinsey estimates that AI-related data center infrastructure investment could reach between $5.2 trillion and $7.9 trillion globally through 2030, with global data center capacity demand expected to nearly triple over the same period. This is not incremental growth; this is a wholesale transformation of the world's energy and infrastructure systems.

How to Identify the Companies Solving AI's Infrastructure Crisis

• Power Generation: Companies developing nuclear energy production, renewable power systems, and grid infrastructure to supply the massive electricity demands of AI data centers.
• Cooling Technology: Manufacturers of advanced cooling systems, liquid cooling solutions, and thermal management equipment designed to prevent AI servers from overheating.
• Semiconductors and Memory: Chip designers and manufacturers producing processors, memory modules, and specialized silicon for AI workloads beyond traditional CPUs and GPUs.
• Electrical Equipment: Makers of transformers, switchgear, power distribution systems, and other grid infrastructure components required to deliver power reliably to data centers.

The RACK ETF tracks the MarketVector Data Center Supply Chain Index (MVRACK), a modified float-adjusted capitalization-weighted index designed to provide exposure to U.S.-listed companies across the data center supply chain. The index methodology ensures that only companies with genuine exposure to AI infrastructure buildout are included, filtering out companies that may have tangential involvement.

Why Are Simultaneous Bottlenecks Creating an Investment Opportunity?

The critical insight is that AI infrastructure faces simultaneous constraints across multiple domains. It is not just a power problem or just a cooling problem or just a semiconductor shortage. All of these constraints exist at the same time, and they are all limiting factors on how quickly hyperscalers can expand capacity. This creates a unique investment thesis: the companies that can solve these bottlenecks may become among the primary beneficiaries of the next phase of AI investment.

"AI demand is creating simultaneous bottlenecks across chips, memory, power and cooling infrastructure. We believe the companies solving those constraints may be among the primary beneficiaries of the next phase of AI investment, and RACK is designed to provide targeted exposure to that buildout," said Frasse.

Nick Frasse, Product Manager, VanEck

VanEck's decision to launch this ETF reflects a broader recognition that infrastructure investing has become as important as software and chip investing in the AI era. The firm has a history of identifying structural shifts early, having pioneered access to international markets in 1993 and launched the first exchange-traded funds in 2006. The RACK ETF builds on VanEck's existing thematic lineup, which includes funds focused on semiconductors, uranium and nuclear energy, and space technology.

For investors and industry observers, the message is clear: the next trillion-dollar opportunity in AI may not be in the software layer or the model layer, but in the physical infrastructure required to make those models run. As hyperscalers continue to pour hundreds of billions into data centers, the companies supplying power, cooling, semiconductors, and electrical equipment will be the ones capturing a significant portion of that spending.