AI Data Centers Face Power Bottlenecks, Driving Infrastructure Investment - Episode Hero Image

AI Data Centers Face Power Bottlenecks, Driving Infrastructure Investment

Motley Fool Money · · Listen to Original Episode →
Original Title:

TL;DR

  • The escalating demand for AI-driven data centers is creating a power and real estate bottleneck, forcing hyperscalers to secure energy sources directly, as evidenced by Microsoft and Amazon's nuclear power initiatives.
  • Data center vacancies are at historic lows, with major customers pre-leasing capacity years in advance, indicating that current operational facilities are insufficient for AI's burgeoning computational needs.
  • Global energy consumption for data centers is projected to double by 2030, with AI-based electricity usage in the U.S. expected to rise 133%, highlighting a critical need for energy infrastructure investment.
  • Companies like HP Enterprise, despite being valued like legacy businesses, are poised to benefit significantly from the AI data center boom by providing high-powered computing systems and networking infrastructure.
  • Real estate investment trusts (REITs) such as Digital Realty Trust, Equinix, and Iron Mountain are diversifying into data center development, capitalizing on the physical asset demand alongside specialized operators.
  • Ancillary businesses in HVAC, civil engineering, and electrical systems, like AAON, Sterling Infrastructure, and Comfort Systems, are quietly benefiting from the data center build-out by providing essential infrastructure and cooling solutions.
  • The increasing reliance on data centers for AI necessitates specialized infrastructure, creating investment opportunities in companies providing critical components like modular nuclear reactors and advanced electrical distribution systems.

Deep Dive

The current build-out of data centers, driven by the immense power demands of artificial intelligence, is not an oversupply issue but a critical bottleneck in energy and real estate. While demand for AI-driven computing power continues to outstrip existing capacity, the true challenge lies in securing sufficient and reliable energy sources, making power availability the new limiting factor in data center expansion. This dynamic creates significant investment opportunities beyond direct data center operators, extending to energy providers and specialized infrastructure companies.

The exponential growth in data center capacity is primarily fueled by AI's need to process and train complex models, leading to record-low vacancy rates and pre-leasing of future space. This surge in demand is straining existing power grids, prompting major tech companies like Microsoft, Amazon, and Google to secure their own energy solutions, including investments in nuclear power and partnerships with energy providers. The International Energy Agency projects global data center energy consumption to double by 2030, highlighting the scale of the challenge and the imperative for new energy infrastructure. This shift in energy sourcing strategies signifies a fundamental change, where direct control over power generation is becoming as crucial as physical location.

The implications for investors are multifaceted, extending beyond direct data center operators and Real Estate Investment Trusts (REITs) like Digital Realty Trust (DLR), Equinix (EQIX), and Iron Mountain (IRM). Companies providing the "picks and shovels" for this build-out are poised for significant growth. HP Enterprise (HPE) is a key player, supplying high-powered computing systems essential for AI infrastructure, yet it trades at a valuation that appears to undervalue its AI exposure. Similarly, companies specializing in mechanical, electrical, and cooling systems, such as Aon (AAON) with its thermal cooling solutions, Comfort Systems (FIX) for HVAC, and M-Core (EME) for electrical and plumbing, are quietly benefiting from the demand for specialized data center infrastructure. Engineering and design firms like Sterling Infrastructure (STRL) and Stantec (STN) are also critical, providing design and construction services. Even established industrial conglomerates like Schneider Electric (trading OTC as SBGSY), known for its Square D brand in electrical equipment, are seeing significant growth from data center sales, offering a diversified play on the underlying power infrastructure.

Ultimately, the data center expansion narrative has evolved from a focus on physical space to a critical dependency on energy. Companies that can provide reliable, scalable, and cost-effective power solutions, alongside the specialized infrastructure to support them, represent the most compelling investment opportunities. The current market appears to be underestimating the strategic importance of power and the diverse range of companies enabling this essential build-out.

Action Items

  • Analyze 5-10 data center projects: Quantify energy consumption impact and identify potential power supply bottlenecks.
  • Audit 3-5 data center construction companies: Assess their capacity to manage specialized infrastructure needs (e.g., HVAC, electrical, concrete).
  • Track 3-5 REITs focused on data centers: Measure their pre-leasing rates and capacity expansion plans against current demand trends.
  • Evaluate 2-3 hyperscalers' energy sourcing strategies: Identify their reliance on traditional power versus investments in modular reactors or renewables.
  • Measure correlation between data center build-out and local electricity costs: Analyze impact across 3-5 affected regions.

Related Episodes

Key Quotes

"Artificial intelligence requires massive amounts of data center capacity to not only handle the queries that are being sent by you and I and enterprises to LLMs, but also to help iterate and constantly train those same models."

Anders Bylund explains that the demand for data centers is driven by the computational needs of artificial intelligence. This includes processing user queries and the ongoing training of AI models, highlighting the fundamental role of data centers in AI development.

"The IEA projects that global energy consumption for data centers alone will double by 2030. And in the United States, the Pew Research Center projects that AI-based electricity usage will rise 133% in that same time period."

Dan Caplinger points out the significant energy implications of data center growth. The projections from the IEA and Pew Research Center underscore the escalating demand for electricity driven by data centers and AI, indicating a substantial future need for power infrastructure.

"So what a lot of hyperscalers are doing is they're taking control of their own destiny and building their own energy sources. Building or buying, I should say. For example, recently Microsoft has restarted or signed a deal to restart Three Mile Island with the kind of traditional reactors that are already there, but they're going to get replaced over time with the smaller modular reactors. So nuclear power in a Microsoft sense."

Anders Bylund describes how major tech companies, or hyperscalers, are proactively securing their energy supply. He uses Microsoft's involvement with nuclear power, including plans for modular reactors, as an example of this trend to ensure sufficient and reliable energy for their data centers.

"Power is the new real estate... for the people who are investing and trying to make AI a reality, it is kind of the real estate that you need. And it is, I think, the bottleneck that investors are missing, which is the energy consumption that is necessary."

Dan Caplinger emphasizes the critical role of energy in the data center build-out. He argues that access to power is now the primary constraint, akin to real estate in previous growth cycles, and that this energy bottleneck is a key factor often overlooked by investors.

"When you look at the top 500 list of supercomputers these days, in the top 10, about six of them, I believe, were built by HPE in the last update in November. And these guys know how to build really high-powered systems that comes in handy at this point."

Anders Bylund highlights HP Enterprise's (HPE) significant contribution to high-performance computing. He notes that HPE builds a substantial portion of the world's top supercomputers, demonstrating their expertise in creating the powerful systems required for AI and data center operations.

"Sterling has also moved into providing, you know, concrete pads and design work on data center installations that they've also done work on some of the systems that go into them and just the project management to get these data centers completed."

Dan Caplinger explains how Sterling Infrastructure (STRL) has expanded its services to support data center construction. He details their involvement in crucial aspects like site preparation, design, system integration, and overall project management, showcasing their role in the physical build-out.

Resources

External Resources

Books

  • "House of Cards" - Mentioned in relation to a quote about power being the new real estate.

People

  • Anders Bylund - Fool analyst discussing data centers.
  • Dan Caplinger - Fool analyst discussing data centers.
  • Emily Flippen - Host of Motley Fool Money.

Organizations & Institutions

  • CPRE - Released reports noting low data center vacancies.
  • IEA (International Energy Agency) - Projects global energy consumption for data centers will double by 2030.
  • Pew Research Center - Projects AI-based electricity usage will rise 133% by 2030 in the US.
  • Microsoft - Signed a deal to restart Three Mile Island for nuclear power, plans to install smaller modular reactors.
  • Amazon - Bought a nuclear-powered campus with plans to install modular mini reactors.
  • Google - Signing deals with independent companies like Nextera Energy for modular nuclear reactors; has a data center in Kiruna powered by a local waterfall.
  • Nextera Energy - Supplier of modular nuclear reactors.
  • Hewlett Packard (HP) - Former conglomerate that split into two parts.
  • HP Enterprise (HPE) - Giant in building high-powered computer systems for data centers; acquired Juniper Networks.
  • Juniper Networks - Acquired by HPE, providing network infrastructure for data centers.
  • Digital Realty Trust (DLR) - REIT concentrating on the data center opportunity.
  • Equinix (EQIX) - REIT diversifying into data centers.
  • Iron Mountain (IRM) - REIT diversifying into data centers.
  • American Tower (AMT) - REIT that concentrated on cell towers for 5G wireless networks and is diversifying into data centers.
  • AON Inc (AAON) - Custom HVAC manufacturing company that acquired Basics, which manages thermal cooling for hyperscalers.
  • Basics - Company acquired by AON Inc, manages thermal cooling for hyperscalers.
  • Sterling Infrastructure (STRL) - Provides concrete pads, design work, and systems for data center installations.
  • Comfort Systems (FIX) - Company with expertise in HVAC applications and data center refrigeration.
  • M Core (EME) - Company with exposure to electrical and plumbing for data centers.
  • Stantech (STN) - Canadian design-focused engineering consultant company with exposure to data center and energy projects.
  • Schneider Electric - French company that manufactures electrical gear under the Square D name; makes power panels and switches for data centers.
  • Square D - Brand owned by Schneider Electric, found in hardware stores and used for data center electrical components.
  • Eaton - Arch rival of Schneider Electric.
  • The Motley Fool - Organization hosting the podcast.

Websites & Online Resources

  • drinkag1.com/fool - Website for AG1, offering a discount for a limited time.

Other Resources

  • Artificial Intelligence (AI) - Requires massive amounts of data center capacity for queries and training models.
  • Data Centers - Discussed as a primary subject, with power and real estate being key opportunities.
  • Supercomputers - HPE builds high-powered systems for data centers, with many on the top 500 list built by HPE.
  • Real Estate Investment Trusts (REITs) - Companies concentrating on data center real estate.
  • Modular Mini Reactors - Technology Amazon plans to install on its nuclear-powered campus.
  • Smaller Modular Reactors - Technology Microsoft plans to use at Three Mile Island.
  • HVAC (Heating, Ventilation, and Air Conditioning) - Custom systems for cannabis greenhouses and thermal cooling for hyperscalers.
  • Picks and Shovels Plays - Underrecognized businesses that help make data centers a reality.
  • Value Chain - Investment opportunities at different points of the data center buildout.
  • Rule Breaker - Category of stocks on the Motley Fool scorecard.
  • Hidden Gem - Category of stocks on the Motley Fool scorecard.

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.