3-Line Briefing
- According to Bloomberg data, electricity rates in areas with dense data center clusters in the U.S. have risen 267% over the past five years, and 13 of the 50 U.S. states now fall within the direct impact zone of AI-driven power demand growth.
- Attention from major investors (smart money) on Wall Street is shifting from semiconductor design and equipment stocks (tickers) like Nvidia and Broadcom toward power generation and transmission/distribution equipment stocks (tickers) that actually supply the electricity to run those chips.
- The basis for investment decisions is shifting from "who makes the faster chip" to "who can supply the electricity to power that chip."
What's Changing
What this statistic really signals is that the AI semiconductor cycle has moved past the stage of competing on computing performance and into a power infrastructure bottleneck stage. The electricity consumption of a single data center has grown to a scale comparable to that of an entire small city, and as this demand concentrates in specific regions, electricity rates are rising in line with basic market logic. Semiconductors only need to go through the foundry process once ordered, but power generation facilities take years for permitting and construction. This time lag is the key reason Wall Street is now focusing on power equipment more than semiconductors.
Whether it's gas turbines or nuclear power plants, new power generation facilities require a lead time of at least several years from planning to completion. Hyperscalers' data center expansion plans, by contrast, are already finalized and underway. In a period where supply cannot keep pace with demand, both the utilization rates and rate-negotiating power of companies that already own power generation and transmission/distribution equipment rise simultaneously. From the perspective of grid operators, this is a phase in which order backlogs convert into actual utilization gains — and there is, again, a time lag before that conversion translates into margin improvement.
Numbers in Context
A 267% increase spread across 13 states shows that this is not a temporary bottleneck confined to a few regions but a structural trend. That said, this figure is limited strictly to areas near data centers and should not be equated with the national average electricity rate. What the market has not yet fully priced in is how much of this rate increase is actually being passed through to hyperscalers' data center operating costs, and whether that cost will in turn spread into higher rates for consumers and industrial users.
Winners and Losers
- Power generation equipment and turbine manufacturers: As new orders for gas turbines and power generation equipment increase, the pricing power of companies that have built up years' worth of backlog strengthens.
- Transmission/distribution grid and substation equipment companies: As demand grows for expanding transmission lines and transformers to supply power to data centers, related orders have room to expand.
- Nuclear power-related companies: Given that data centers require stable, round-the-clock power, demand for nuclear power purchase agreements (PPAs) could come into greater focus.
- Domestic (Korean) power equipment and heavy electrical machinery makers: They may indirectly benefit from the surge in U.S. grid expansion, though actual order wins need to be confirmed through individual company disclosures.
- Semiconductor design and foundry stocks (tickers): If the power bottleneck slows the pace of data center expansion itself, there is a risk that growth in demand for new chips could decelerate.
Risk Check
- Power infrastructure-related stocks already have substantial expectations priced into their valuations, so there is a risk of a pullback if actual order and utilization-rate indicators fail to follow through.
- Expansion of power generation facilities is subject to policy variables such as permitting, environmental regulations, and local resident opposition, so there is a possibility that plans may not proceed as scheduled.
- If rising electricity rates translate into a heavier cost burden for data center operators, it could instead become a headwind that slows the pace of new investment.
- The 267% figure is limited to areas with dense data center clusters, so it would be premature to directly link it to an across-the-board earnings improvement for the power infrastructure industry sector.
Bottom Line
The fund flow shifting from semiconductors to power is a reasonable signal that the AI infrastructure cycle has entered its next phase, but given the long lead times for power generation facilities and policy variables, the timing and pace at which this translates into actual profit need to be verified separately through individual order disclosures and utilization-rate indicators.
This article is content automatically summarized and analyzed based on the original news report. View original (Maeil Business Newspaper, Securities)
