CONTRIBUTORS
Christian Correa, CFA
President & Chief Investment Officer
Franklin Mutual Series
Artificial intelligence’s (AI’s) hype phase may be over. With tech firms investing tens of billions of dollars in AI and new competitors continuing to emerge, a more subdued investment phase may eventually take hold.
Nonetheless, the potential for AI to have a transformative impact continues to draw interest and investment. Not only is this investment in new chips and software, but also in the old economy companies that build the infrastructure to support tech’s AI ambitions. And as AI models get cheaper and more users learn how to get useful results from them, we could see usage, and the resources needed to support AI, surge. So far, broad-based adoption seems far off. But like past booms, be it the railroad, mainframe computer or internet frenzies, AI’s full potential may become much clearer over the next decade.
With US companies still spending lavishly to fill a shortage of computing power, data centers and energy, value investors need to remain picky and nimble, in our opinion, when looking to take advantage of potentially cheaper AI opportunities.
Spend, spend, spend?
The recent emergence of a new AI competitor that needs fewer chips, less capital and less energy than those models the major US tech firms run has raised questions. Can the boom in AI-related capital spending continue at its current pace without a pickup in AI users and profits?
Major tech companies like Google, Meta, Amazon, Oracle and Microsoft are spending vast sums of money on AI-related infrastructure. Their fourth-quarter earnings announcements showed no signs of a pullback in investment in servers, routing and optical equipment, load balancing and latency in addition to the chips that power the models. All this spending has translated into a corresponding increase in high-tech industrial production in recent years. (See Exhibit 1.)
Exhibit 1: US Tech Spending Boosts Tech Industrial Production
US Tech Industrial Production Surges
January 2014-January 2025
Seasonally Adjusted
Source: US Federal Reserve.
However, like in past tech spending cycles, be it the mainframe expansion in the 1960s and 1970s, personal computers in the 1980s or the internet boom of the late 1990s, how people ultimately use this computing power and the new services and applications that spring up to support it may not fully manifest for a decade or so. According to the US Census, only 8% of US households owned a personal computer (PC) in 1984. It took until about the year 2000 before half of US households had them and another 10 years before nearly 80% of the population had a PC.1 It was their increasing usefulness that spurred greater adoption. Big US tech companies are spending billions profitlessly on AI today in hopes of bigger future rewards.
As value investors, we see few investment opportunities in this space currently—the semiconductor and hardware companies have run hard over the past few years—but years of research and experience suggest to us that opportunities for AI-related value investments may arise either as current darlings stumble, or as out-of-favor firms find innovative ways to tap into AI’s vast potential. We believe research and an eye on valuations will be crucial.
Building out the infrastructure
As the spending continues for now, we see an ongoing need for larger data centers to run the AI models and process the expected jump in data volumes. Some existing data centers are being repurposed to support AI, while others are being built from scratch in places like Northern Virginia, known as “data center alley.” AI optimism has led US spending on data centers to climb sharply over the past year as shown in Exhibit 2.
Exhibit 2: The US Is Building More Data Centers
US Data Center Construction Jumps
January 2014-November 2024
Millions of Dollars, Not Seasonally Adjusted
Source: US Census Bureau.
This construction is providing many traditionally value industrial companies with new economy opportunities. AI data centers are more power and heat intensive than standard data centers. The large-language AI models are being run nonstop in the training phase, as they are fed information. That leads to greater electricity consumption, which generates heat, and requires not only more physical space, but also electrical and mechanical equipment to run all the servers and tech gear as well as water and equipment to keep the data centers cool and reduce the possibility of overheating or power disruptions.
The servers, for instance, need to be cooled locally, so that the racks they sit on may have coolants running through them, with chillers sitting outside the data centers. The air also must be kept cold, requiring HVAC systems. Meanwhile, keeping the data centers running requires power distribution systems, switches and gear to ensure power supply is uninterrupted. And the whole data center needs to be tied back into the electrical grid with transmission lines or powered by mobile generation solutions. (See image below.)
Overhead View of a Data Center
Source: Adobe Stock.
Overall, spending on the required industrial equipment needed to run new data centers could total US$250 billion by the end of the decade, consulting firm McKinsey estimates.2 While many industrial stocks may already reflect the optimism about all this AI data center spending, we believe there could be places where both large- and small-cap value investors may still find overlooked opportunities.
A power revolution
Powering all this tech and industrials equipment also requires more and more power. A utility sector that has not seen load growth in nearly 40 years is at the beginning of a major power revolution, in our view. Overall, the US Department of Energy and the Berkeley Lab estimate that US data center electricity demand should grow at an increasing rate over the next few years, reaching 6.7%–12% of total US power consumption by 2028 compared with 4.4% in 2024.3
After years of flat or shrinking power demand around the developed world, the staid utilities sector may need to spend more on both power generation and transmission to meet the growing demand from AI data centers. And currently, access to sufficient power has constrained some data center construction, which we believe means there is pent-up demand still to be unleashed.
Bain & Co., a US consultancy, forecasts that utilities will need to increase annual energy generation by as much as 26% above 2023 levels by 2028 to meet projected demand, well above the 5% boost seen in the prior two decades.4 We have seen US utility industry capital expenditures begin to rise to meet the growing demand for US electricity. (See Exhibit 3.)
Exhibit 3: US Utilities Spend More to Meet Growing Demand
US Utility Capital Expenditures Climb
2014-2026
Billions of US Dollars
Source: Edison Electric Institute. E=estimated. There is no assurance that any estimate, forecast or projection will be realized.
All this new power demand will require a modernized electrical grid, greater generation facilities and equipment like gas turbines and green energy alternatives given grid capacity constraints. With power demand likely to increase and supply likely to rise to meet it, we expect to see growth for the sector, for the first time in decades, over the next few years. The boring old utilities sector could start to see growth, providing a necessary catalyst for utility sector multiple expansion beyond the increases we saw in 2024.
A decade’s wait?
While all this money goes into building the infrastructure to support AI, the meaningful use cases seem way off. We expect that AI’s full potential may become clearer over the next decade, as the tech booms of the past have suggested. Whether AI investment can be made more cheaply and with less tech and fewer data centers is now an open question.
Nonetheless, as value investors, the rise in AI is not just a growth and technology story. Realizing this new technology’s potential still needs more traditional electrical equipment, cooling systems, and power generation and transmission over time. Recent market gyrations may provide value investors with potential opportunities to seek to take advantage of this growth as the old economy meets the new; new opportunities may also arise as AI becomes more widely adopted. In the meantime, be mindful of the valuations.
Endnotes:
- Source: “Computer and Internet Use in the United States: 2016.” US Census. August 2018.
- Source: “AI power: Expanding data center capacity to meet growing demand.” McKinsey & Company. October 29, 2024.
- Source: “2024 United States Data Center Energy Usage Report.” Berkeley Lab. December 2024.
- Source: “Utilities Must Reinvent Themselves to Harness the AI-Driven Data Center Boom.” Bain & Company. October 2024.
WHAT ARE THE RISKS?
All investments involve risks, including possible loss of principal.
Equity securities are subject to price fluctuation and possible loss of principal.
Large-capitalization companies may fall out of favor with investors based on market and economic conditions. Small- and mid-cap stocks involve greater risks and volatility than large-cap stocks.
Value securities may not increase in price as anticipated or may decline further in value. Growth or value as an investment style may become out of favor, which may have a negative impact on performance.
Active management does not ensure gains or protect against market declines.
Investment strategies which incorporate the identification of thematic investment opportunities, and their performance, may be negatively impacted if the investment manager does not correctly identify such opportunities or if the theme develops in an unexpected manner. Focusing investments in technology-related industries, carries much greater risks of adverse developments and price movements in such industries than a strategy that invests in a wider variety of industries.
Securities issued by utility companies have been historically sensitive to interest rate changes. When interest rates fall, utility securities prices, and thus a utilities fund’s share price, tend to rise; when interest rates rise, their prices generally fall.
Any companies and/or case studies referenced herein are used solely for illustrative purposes; any investment may or may not be currently held by any portfolio advised by Franklin Templeton. The information provided is not a recommendation or individual investment advice for any particular security, strategy, or investment product and is not an indication of the trading intent of any Franklin Templeton managed portfolio.