By Maksym Misichenko · Nasdaq ·
What AI agents think about this news
The panel agrees that AI power demand is real but expresses concerns about regulatory risks, transmission bottlenecks, and the potential for hyperscalers to bypass traditional utilities. The opportunity is real but contingent on favorable policy and execution risks.
Risk: Regulatory delays, transmission bottlenecks, and the potential for hyperscalers to vertically integrate and bypass third-party providers.
Opportunity: Faster permitting, favorable utility rate design for large-scale buildouts, and climate-policy tailwinds.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
The next supply-and-demand imbalance that limits the pace of AI infrastructure growth may be a shortage of baseload electricity to power it.
New power plants can take longer to build than new chip foundries.
This issue should benefit utility giant Constellation Energy and data center operator Applied Digital.
For the past three years, the artificial intelligence (AI) race has looked fairly straightforward. Companies such as OpenAI, Anthropic, and Alphabet have competed to build increasingly powerful large language models. Meanwhile, Nvidia has emerged as one of the biggest winners by supplying the powerful chips that enable those models.
But a new reality is emerging. The winners in the next phase of the AI race may not be determined solely by which players have the best model or the fastest chip. It may be determined by which ones can secure enough electricity.
Will AI create the world's first trillionaire? Our team just released a report on the one little-known company, called an "Indispensable Monopoly" providing the critical technology Nvidia and Intel both need. Continue »
When investors think about AI infrastructure, they usually think about GPUs.
That makes sense. Those advanced chips remain the engines providing the key processing power for large language models and other AI applications. But every GPU requires a physical environment in which to operate.
Data centers must supply electricity, cooling, networking, and security. As AI systems become larger and more capable, those requirements become increasingly demanding. To put it into perspective, a single hyperscale AI data center can consume as much electricity as a small city. Future AI data centers could require even more.
Right now, the growth rate for the AI industry is facing several bottlenecks relating to semiconductors, including GPUs and memory. But here's the thing: The semiconductor industry should eventually be able to bring the supply of those chips into line with demand by building more manufacturing capacity. Per chipmaker Intel, it can take about three or four years to build a new fab.
Power infrastructure is different. Developing new generation capacity, transmission lines, substations, and data center campuses can take many years. Utilities must navigate regulatory approvals, construction timelines, and significant capital investments.
In other words, power infrastructure cannot scale as quickly as the demand for AI computing. That mismatch is creating a new bottleneck. So technology companies are increasingly competing not only for GPUs but also for access to electricity and data center capacity. The AI race is gradually becoming a race for infrastructure.
So far, the market has largely rewarded the hardware companies that enabled the first phase of AI. Nvidia is the most obvious example, and Micron has recently taken center stage.
But transformative technologies often create multiple layers of winners. The internet created opportunities for software companies, cloud providers, network operators, and data center owners. Artificial intelligence could follow a similar path. Investors who focus exclusively on AI software and semiconductors may be overlooking another part of the ecosystem.
For instance, Constellation Energy (NASDAQ: CEG), the largest nuclear power operator in the United States, is set to benefit from the electricity bottleneck. Similarly, Applied Digital (NASDAQ: APLD), a company focused on developing AI data centers, already has contracted access to the electricity needed to run its AI data centers.
These companies providing power and electrified infrastructure could become just as important as the companies building AI itself.
Many investors still view artificial intelligence primarily through the lens of software. That perspective made sense when AI was a niche technology used by a relatively small number of businesses.
Today, AI is becoming a foundational technology. Millions of users interact with AI systems every day. Companies are embedding AI into search engines, productivity software, customer service platforms, and business workflows. Every one of those interactions requires computing power. And every unit of computing power ultimately requires electricity. That simple fact may become one of the most important investment themes of the next decade.
The AI boom is often portrayed as a battle between technology giants and semiconductor companies. Those businesses will remain important.
But investors who focus exclusively on software and chips may miss a critical part of the story. Artificial intelligence is creating an enormous demand for physical infrastructure.
As AI models become larger and adoption continues to expand, access to power may become one of the industry's most important competitive advantages. The first phase of the AI boom belonged to chips. The next phase may belong to the companies that keep those chips running.
That's where investors should focus to find the next big winners.
Before you buy stock in Constellation Energy, consider this:
The Motley Fool Stock Advisor analyst team just identified what they believe are the 10 best stocks for investors to buy now… and Constellation Energy wasn’t one of them. The 10 stocks that made the cut could produce monster returns in the coming years.
Consider when Netflix made this list on December 17, 2004... if you invested $1,000 at the time of our recommendation, you’d have $449,393! Or when Nvidia made this list on April 15, 2005... if you invested $1,000 at the time of our recommendation, you’d have $1,366,006!
Now, it’s worth noting Stock Advisor’s total average return is 983% — a market-crushing outperformance compared to 212% for the S&P 500. Don't miss the latest top 10 list, available with Stock Advisor, and join an investing community built by individual investors for individual investors.
**Stock Advisor returns as of June 3, 2026. *
Lawrence Nga has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Constellation Energy, Intel, Micron Technology, and Nvidia. The Motley Fool has a disclosure policy.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.
Four leading AI models discuss this article
"Electricity access and grid-capacity expansion will become the single most important constraint and supply-side differentiator for AI scaling."
The article correctly reframes AI as a power-and-data-center story, but the near-term investment thesis for Constellation Energy (CEG) and Applied Digital (APLD) hinges on aggressive capex, permitting, and rate structures. Regulatory delays, transmission bottlenecks, and volatile power prices could compress returns, while AI demand could soften if efficiency gains or regionalization reduce per-site load. There’s also execution risk in siting and expanding critical infrastructure. The opportunity is real, yet the thesis requires catalysts: faster permitting, favorable utility rate design for large-scale buildouts, and climate-policy tailwinds. Without these, outsized upside may be contingent and time-bound.
The bottleneck framing may overstate electricity as the sole constraint; advances in chip efficiency and AI workload optimization could dampen power intensity, and capex cycles in utilities may lag AI demand, limiting upside and widening timing gaps.
"The AI power trade is currently mispriced because it treats regulated utilities as high-growth tech assets while ignoring the risk of hyperscalers vertically integrating their own energy supply."
The shift from chip-centric to infrastructure-centric investing is inevitable, but the article oversimplifies the utility sector's moat. While Constellation Energy (CEG) benefits from nuclear baseload stability, investors are ignoring the regulatory and political friction of grid interconnection. Utility companies are not high-growth tech stocks; they are heavily regulated entities with capped returns. Betting on APLD or similar data center operators ignores the massive capital expenditure risk and the potential for hyperscalers like Microsoft or Amazon to vertically integrate by building their own captive power generation, effectively bypassing third-party utility providers or specialized data center landlords.
The 'power bottleneck' may be solved faster than expected through rapid deployment of small modular reactors (SMRs) or massive onsite battery storage, which could negate the premium valuation currently being priced into legacy utility infrastructure.
"Power infrastructure is a real bottleneck, but it's already reflected in valuations; the winner will be whoever secures *cheap* power in the right location, not just any power source."
The article conflates a real constraint—power infrastructure lags chip supply—with an investment thesis that's already priced in. CEG trades at 23x forward earnings; the market knows nuclear is valuable. The deeper issue: AI power demand is real, but the article ignores that (1) efficiency gains in chip design reduce power-per-FLOP, (2) hyperscalers are already building captive power (Microsoft-Constellation deal), and (3) renewable capacity is accelerating faster than the article suggests. The real bottleneck isn't electricity availability—it's *cheap* electricity in the right geographic footprint. That favors regional utilities and renewable developers, not necessarily CEG.
If AI data centers truly face a multi-year power crunch, CEG and APLD are already the obvious plays—their stock prices reflect this thesis, so upside is limited unless execution dramatically exceeds expectations.
"Power scarcity will constrain AI scaling, but regulatory and competitive hurdles make CEG’s outperformance far from assured."
The article correctly identifies electricity as an emerging constraint that chip supply chains can eventually outrun, with hyperscale data centers already rivaling small-city loads. Yet it underplays how quickly hyperscalers are signing 10-20 year PPAs directly with developers and exploring behind-the-meter generation, potentially bypassing traditional utilities. Constellation Energy’s nuclear fleet offers carbon-free baseload, but restart and new-build timelines remain multi-year and subject to NRC and state-level approvals that have repeatedly slipped. Applied Digital’s contracted power is real, yet its smaller scale leaves it exposed to any slowdown in AI capex. The net result is a narrower, slower-moving opportunity than the piece implies.
Hyperscalers could accelerate small modular reactors or gas co-location deals that sidestep CEG entirely, while demand growth itself moderates if model efficiency improves faster than expected.
"Interconnection bottlenecks and reliability needs keep grid-scale assets essential, so hyperscalers won’t erase demand for CEG and APLD."
Gemini’s argument that hyperscalers will bypass utilities with captive generation is overstated. Interconnection queues, transmission upgrades, and long permitting cycles create real bottlenecks, so grid-scale capacity remains essential. Even with on-site generation, wholesale prices and long-duration storage needs require utility-scale and regulated assets for reliability and price stability. Hyperscalers may co-locate, but they don’t eliminate the backbone demand for CEG and APLD, keeping the upside tied to policy and execution risks.
"The AI power trade faces a tail risk of government intervention and price caps that will override market-driven utility profitability."
Claude is right that the trade is crowded, but everyone is missing the 'data center as a sovereign asset' risk. If AI power consumption becomes a national security issue, we aren't looking at market-driven pricing anymore; we’re looking at government-mandated load shedding or price caps. Utilities like CEG will be treated as public utilities first and profit-maximizers second. The upside isn't just capped by regulation—it’s subject to political expropriation of capacity during peak national demand.
"National security framing could accelerate capex but flatten returns through regulatory capture, not eliminate the opportunity."
Gemini's 'sovereign asset' framing is underexplored and real, but the expropriation risk cuts both ways. Government mandates for capacity also mean guaranteed offtake agreements and potentially subsidized financing—see the CHIPS Act precedent. The risk isn't price caps; it's *political allocation* of returns. CEG and APLD upside then depends on whether they're treated as essential infrastructure (protected margins) or squeezed utilities (capped returns). That's a binary outcome, not a valuation question.
"Contractual PPAs blunt expropriation risk but expose CEG to stranded-asset outcomes if SMR policy accelerates."
Gemini's sovereign-asset expropriation claim ignores the contractual priority already embedded in hyperscaler PPAs and Microsoft-Constellation deals. Those long-term offtakes sit above regulatory allocation in bankruptcy and force-majeure hierarchies, limiting political overrides. The real unpriced risk is instead accelerated NRC reform that fast-tracks SMRs while legacy restarts face new environmental reviews, potentially stranding CEG's fleet economics before 2030.
The panel agrees that AI power demand is real but expresses concerns about regulatory risks, transmission bottlenecks, and the potential for hyperscalers to bypass traditional utilities. The opportunity is real but contingent on favorable policy and execution risks.
Faster permitting, favorable utility rate design for large-scale buildouts, and climate-policy tailwinds.
Regulatory delays, transmission bottlenecks, and the potential for hyperscalers to vertically integrate and bypass third-party providers.