Data Centers Need 100+ GW. My Favorite Utility Play Is the One With Real Earnings Power.

The bottleneck is the grid, not the AI narrative

This is not a story about pointing at AI and calling it a winner. The real conflict is simpler: AI needs power, but the grid is struggling to deliver it fast enough. Bank of America sees a U.S. supply gap of more than 100 GW over the next five years, while Goldman Sachs says data-center power demand alone is set to jump from 31 GW in 2025 to 66 GW in 2027. That points to a real infrastructure squeeze, with delays, wait times, and tighter power markets likely to matter more over time.

Utilities already showed how the market reacts to this theme

Investors have already had a chance to test the trade. The utilities group rallied 12% to an all-time high on February 27, then fell 7% from that February peak as macro conditions cooled. That pattern suggests investors like the AI-power idea in principle, but are less willing to pay for it when valuation or rates turn against the sector.

What matters here is earnings visibility. Regulated utilities can turn rising demand into rate base growth and, if regulators allow, steady earnings power. That is a different setup from many AI names that still need to prove monetization. Morningstar still expects 6%-8% average annual earnings growth for the sector, with data centers remaining a meaningful demand tailwind. The key is to focus on companies where that demand can actually flow through to earnings.

Why the utility thesis has real earnings power

This thesis works because the mechanism is straightforward: more load means more wiring, poles, substations, and grid upgrades. In a regulated model, that spending can become durable earnings if it moves through rate base and cost recovery.

How the earnings engine works

The core idea is simple. Grid investment rises with load, and utilities are generally allowed to earn a return on approved regulated assets. That does not require an instant spike in net income. It requires a visible pipeline of spending and a realistic path to regulatory recovery. That is why steady rate base expansion keeps showing up as a core support for utility earnings.

But the main risk is real: this only works if regulators allow cost recovery. If projects get approved but rate increases lag, investors can overpay for growth that never fully shows up in earnings. So the right test is not just whether demand is rising, but whether management can turn that demand into rate base and allowed returns.

The physical bottlenecks are already visible

The shortage is not theoretical. Equipment and construction constraints are starting to show up in the market. Utilities are dealing with large gas turbines largely sold out through 2030, while only 50–60% of data center capacity scheduled for the next one to two years is expected to come online on time. Those delays matter because they keep pressure on grid upgrades, interconnection queues, and project timelines.

There is also a more constructive development for utility investors: larger customers may be asked to bear more of the cost of the upgrades their projects require. States are moving toward a framework where large-load customers pay for the grid infrastructure upgrades their facilities need, and Virginia is part of that shift. That does not guarantee outcomes, but it improves the odds that new spending can land cleanly in rate base.

The bear case is about timing, not logic

The market has already granted utilities some credit for this story. The sector's 3% dividend yield is near multi-decade lows, and Morningstar has warned that returns could disappoint if earnings do not grow fast enough. So the main risks are straightforward:

• rate recovery slips
• capex rises but does not translate into allowed returns
• delay risks keep project timelines messy

If those risks materialize, the stocks can lose appeal quickly because investors would be taking utility risk for less income than before.

My preferred way to play the trend: regulated utilities with direct load exposure

The cleaner play is not the whole utility basket. It is the regulated names where data-center load lands directly on their wires and, ideally, their next rate case. You want companies with regulated rate bases and large regulated customer bases, because that is where rising demand is most likely to become rate base growth and steady returns. That is different from owning a broader power complex or a speculative grid story; regulated utilities do not need to make a commodity-price call.

Focus on markets where the grid is already tight

I would lean toward utilities in areas where the system is already showing strain, not where new power is assumed to arrive on schedule. Goldman sees elevated reliability risks in the Mid-Atlantic and Mid-Continent as data-center demand rises. In those markets, the likely winners are the utilities that can support that load with substations, lines, interconnection work, and other regulated grid assets. The same regulatory shift that asks large customers to pay for the grid infrastructure upgrades their facilities require also improves the odds that that investment can support returns.

What to watch next

Keep the framework simple. The best way to play this trend is through regulated utilities that have visible load growth, visible rate-base upside, and a credible path to cost recovery. From there, the important signals are practical: project approvals, equipment lead times, interconnection delays, and whether utility guidance keeps tying new investment to earnings rather than just to demand.