Our latest Disruptor Series episode digs deep into the enablers of the AI innovation wave.
The major AI players—including Google, Microsoft and Amazon—have dominated investors’ conversations, media coverage, and indices for some time. But AI requires heavy-duty processing that’s housed in data centers, and data centers crave power that has to be generated, transmitted and distributed to meet that need.
Each of these connections is part of the AI supply chain, and investors must understand each link to put their arms around the full range of AI opportunity. Those opportunities often cluster around bottlenecks in the chain, as power demand rises and forces the energy industry to balance the old guard with a “new guard.”
This was the theme of a wide-ranging discussion in our most recent AllianceBernstein (AB) Disruptor Series episode. A brief summary of the discussion follows.
Power Generation—Creating the Juice for AI
Many parts of the AI supply chain depend upon elements. Two are neodymium, a rare earth element used in magnets for power turbines, and uranium, a less-rare element that’s critical for nuclear power. Supply-chain bottlenecks exist for each in sourcing and processing—with geopolitical risk based on location. The race to secure reliable access gives rise to investment opportunities.
Nuclear power offers base-load power for AI data centers—generating 24/7. In the US, the existing large power plants are seeing their lives extended—and in some cases, AI companies seek to plug their data centers directly into plants. Small modular reactors (SMRs) may be the faster and cheaper path to more capacity, but the US hasn’t built such facilities in a long time and is trying to reverse course from decades of eschewing nuclear power. It will take time to loosen the bottleneck, creating potential for investors in firms with existing large-scale nuclear facilities and those exploring SMRs.
Gas turbines enjoy cheap fuel and are effective, but power demand has been flat for decades, so the US hasn’t built many turbines. The capacity of existing producers is booked and funded for five years, stoking demand for companies that can produce excess supply. In renewables, solar and wind have become very cost-competitive and make up the vast majority of new supply. But they’re intermittent, which isn’t optimal for “always on” data centers, so renewable alone isn’t the answer.
The bottom line: if demand rises as expected, it will be all hands on deck for power generation—each source with its own opportunities and challenges and part of the ultimate solution.
Utilities: Two Distinct Opportunities
For utilities, which own the equipment and transmit and distribute power, capital investment has been feeble given flat power demand. Today, the growth cycle to replace aging infrastructure is running into AI’s wave of power demand, but there are bottlenecks created by affordability considerations.
We see opportunities with two distinct risk profiles. For traditional, vertically integrated utilities under regulation, growth comes from rising demand or equipment replacement. Which firms stand to benefit from power-demand growth? Do they operate in regions like Texas, where it’s growing? Is the need for infrastructure replacement stronger?
Some companies own power-generation assets in deregulated markets that are under the strain of fixed-cost facilities and a period of plummeting energy prices. Because they’re not necessarily looking to build more capacity, they have scarcity value—and efforts to monetize their assets create unique opportunities. For example, a plant might contract with a data center to sell power at a fixed premium price. And if broader power demand goes up, they can win that way.
For investors, the challenge is determining whether the cycle of higher power demand is sustainable and can bolster the revenue picture and whether efficiency can reduce operating expenses.
The Grid: Power Transmission and Distribution
The US power grid hasn’t really changed much in a century: high-voltage lines and big transformers move the power; substations and smaller transformers deliver it to customers.
But the continued migration to renewable power generation is complicating the grid. Add that to the challenges of utility operators, and we see a huge infrastructure opportunity. More complex grids need more-diverse, better-integrated equipment as demand grows and power-generation capacity tries to catch up to power demand.
Key transmission pieces—utility-scale transformers, substations and switching equipment—tend to be dominated by global players. There are also wire and cable providers operating in a highly competitive space. On the distribution side are smaller transformers, switch gear and the equipment directly attached to buildings receiving power.
Given this landscape, one approach investors should consider is emphasizing market leaders in types of equipment where the changing mix of power generation doesn’t have much impact. They don’t care where the power comes from; they’re positioned to win either way.
AI Data Centers: Understanding the “White” and “Gray” Space
In our trip along the AI supply chain, we now arrive at the AI data center—a power-thirsty customer. AI requires enormously powerful computers: tens of thousands of processors working together as a single supercomputer. Smarter AI calls for more-efficient processing and interaction. More-robust processors need special packaging. These needs all create links in the supply chain.
One supercomputer is the size of a refrigerator and consumes as much energy as a big supermarket at its busiest peak. Because these processors run hot, they need sophisticated cooling systems. That takes us from the “white space” of data centers—the server room—to investment opportunities in the “gray space” around server rooms. These “gray” areas are also home to industrial-grade power-regulating equipment. Many firms play a role in building and installing all the parts that keep data centers humming.
It takes a lot of focus to stay on top of the AI evolution—which won’t follow a straight line. It also requires fundamental research to understand how these changes will create winners and losers in the AI supply chain. Those who gain this insight have the chance to tap a massive opportunity.
AB’s Disruptor Series is designed to provide distinctive perspectives on critical issues facing the capital markets today.
