The AI Power Bottleneck Trade — $SEI / $WTS / $BE

Executive Summary

The AI power bottleneck trade is built on a simple but powerful shift: compute demand has moved faster than the electrical system that must support it. For years, investors could analyze AI through the semiconductor supply chain: GPUs, HBM memory, networking, servers, cloud budgets and software adoption. In 2026, the market is increasingly treating power availability as a gating factor. A data center that cannot secure electricity is not a data center; it is a construction project waiting for permission from the grid.

This matters because the physical bottleneck is different from a normal technology cycle. Chip cycles can reset quickly when supply catches up or demand pauses. Grid infrastructure, power plants, substations, transmission upgrades, water systems and cooling retrofits move on slower industrial timelines. If AI demand remains strong, the companies that can help customers secure power, manage heat, reduce water risk and deploy modular solutions may receive more durable investor attention than many second-tier software names that merely attach “AI” to a slide deck.

The three tickers in this report offer three different lenses on the same theme. Solaris Energy Infrastructure is the most direct behind-the-meter power name in the group. Watts Water Technologies is the cleaner industrial infrastructure compounder tied to water and flow-control systems, with AI data centers providing a growth narrative rather than the whole business. Bloom Energy is the highest-profile fuel-cell name in the AI power conversation, but after its enormous repricing it no longer fits neatly inside the small/mid-cap bucket. It is included here because it has become the public-market reference point for how aggressively investors are willing to price fast-deployable power for AI workloads.

The key editorial angle is not “buy these stocks.” The key angle is that Wall Street is repricing a new category of AI beneficiaries: companies that solve the power, water and cooling limits beneath the AI boom. That creates opportunity, but also creates obvious risk. Valuations can detach from fundamentals, contracts can be delayed, grid politics can change, regulators can push back, customers can renegotiate, and the same excitement that lifted the group can reverse if hyperscaler capex expectations cool.

Why This Theme Is Hot Right Now

The immediate reason is that the power issue has moved from specialist industry conferences into the center of market discussion. On June 18, Reuters reported that the top U.S. energy regulator directed regional grid operators to reconsider how they connect very large energy users such as data centers, because server warehouses are straining power grids and pushing U.S. electricity consumption toward record highs. That is exactly the kind of policy-level confirmation investors look for when a theme moves from “interesting” to “structural.”

Separately, Reuters reported earlier in June that U.S. power consumption is expected to reach new highs in 2026 and 2027, driven by AI-hungry data centers and electrification. The EIA projection cited in that report shows U.S. electricity demand moving from a record 4,195 billion kWh in 2025 to 4,271 billion kWh in 2026 and 4,397 billion kWh in 2027. Those numbers are not stock-promotion material. They are the macro foundation under the trade.

The second reason is that AI data centers are not normal office buildings with a few extra servers. They are power-dense industrial assets. Their economics depend on uptime, redundancy, access to fast electricity, cooling performance and the ability to deploy capacity before competitors capture the compute opportunity. If a cloud customer wants capacity now and the grid interconnection takes years, the value of modular generation and behind-the-meter solutions rises dramatically.

The third reason is that the public debate is becoming sharper. Reuters also documented how dozens of fast-tracked, off-grid natural-gas power plants are being proposed or built in the United States to serve individual data centers, sometimes with limited public scrutiny. That is not just an environmental story. It is an investable clue. The market is seeing evidence that data-center developers are willing to find power outside the slowest parts of the traditional utility process. That opens room for companies that can deliver power packages, electrical distribution, modular generation, fuel cells, thermal controls and grid-support systems.

The fourth reason is water. AI infrastructure is increasingly colliding with local resource constraints. Data centers can be more efficient than older designs, and hyperscalers are improving water usage. But the core public concern remains: large compute facilities need energy, cooling and local infrastructure, and the environmental footprint is increasingly visible. This matters for Watts Water because the water side of the AI trade is less crowded than the power side. Investors have already discovered the obvious power names; water and flow-control systems may be a second-order infrastructure angle that still feels under-owned compared with fuel cells, turbines and utilities.

The Core Thesis: AI Has A Physical Bottleneck

The most important shift for investors is conceptual. The first stage of the AI boom was digital. The second stage is physical. The digital stage asked whether large language models could generate demand, whether enterprises would pay for AI, whether cloud providers would increase capex, and whether semiconductor supply could satisfy demand. The physical stage asks a different set of questions: where will the power come from, how fast can it be delivered, who pays for grid upgrades, how much water is available, and what cooling technology can handle dense AI racks?

That creates a broader investment map. Nvidia, AMD, Broadcom, Micron and the server OEMs are still central, but they are not the only gatekeepers. Power suppliers, transformer makers, electrical contractors, gas turbine manufacturers, fuel-cell vendors, water technology companies, cooling specialists, data-center REITs and utilities all become part of the AI supply chain. In other words, AI is starting to look less like a pure software theme and more like a massive industrial construction cycle.

The phrase “power bottleneck” is useful because it avoids overcomplicating the story. The bottleneck has several layers. There is generation: enough electricity must exist. There is transmission: electricity must move to the right location. There is interconnection: a project must receive permission and infrastructure to connect. There is distribution: power must be stepped, routed and protected within the data-center campus. There is thermal management: heat must be removed reliably. There is water or water avoidance: cooling systems must function in a world where local communities increasingly scrutinize water consumption. There is redundancy: downtime is unacceptable.

Each company in this report touches a different part of that chain. Solaris is closest to immediate power deployment. Watts is closer to water infrastructure and flow-control reliability. Bloom is closer to distributed fuel-cell generation. The combined article works because the three names tell a coherent story: AI infrastructure is moving beyond chips and into the real-world systems that make compute possible.

$SEI — Solaris Energy Infrastructure: The Most Direct Power Bottleneck Name

The key fact is scale. Solaris’ April 2026 investor presentation referenced approximately 3.1 GW of secured power generation capacity, including capacity scheduled for delivery through 2029. Its Q1 2026 earnings supplement highlighted a new commercial contract for more than 600 MW plus balance-of-plant services for a minimum term of 10 years, with a five-year option, to support data-center compute needs beginning in late 2026 and scaling through 2028.

That is why $SEI has become more than an oilfield services-adjacent name. The market is no longer valuing it simply as a cyclical energy services company. It is starting to treat Solaris as a behind-the-meter power partner for technology customers that need large, fast, reliable energy blocks. The phrase “behind the meter” matters because it points to power solutions located on or near the customer site, potentially reducing dependence on traditional grid interconnection timelines.

Solaris’ Q1 numbers gave the market hard evidence. The company’s Q1 2026 earnings supplement showed total revenue rising from $126 million in Q1 2025 to $196 million in Q1 2026, while adjusted EBITDA increased from $47 million to $84 million over the same period. That kind of revenue and EBITDA growth gives the AI infrastructure narrative something more valuable than a slogan: operating traction.

The bull case is straightforward. If hyperscalers and other large technology customers need power faster than utilities can provide it, companies with modular generation capacity, electrical distribution capabilities and turnkey project execution become strategically important. Solaris’ long-term contracts improve visibility. The company’s 10-year-plus contract durations suggest customers are not only renting emergency backup equipment; they are committing to multi-year power infrastructure arrangements.

The bear case is also straightforward. Solaris is scaling into a capital-intensive opportunity. The same investor materials reference financing and refinancing considerations to support growth initiatives. Rapid growth in generation capacity can create a powerful earnings ramp, but it can also create balance-sheet risk if capital costs rise, customers delay deployments, equipment availability tightens, or contract economics prove less attractive than investors expect. A small/mid-cap infrastructure company can reprice aggressively when it wins major contracts, but the market will eventually demand evidence that those contracts convert into durable free cash flow.

The other risk is concentration. Large data-center power contracts are attractive, but they may also make the equity story dependent on a small number of major technology customers. If one customer changes deployment timing, project scope or financing terms, the stock can react sharply. For investors, that makes $SEI a high-conviction infrastructure name, not a sleepy utility-like compounder.

Still, as a theme stock, Solaris is hard to ignore. It sits at the point where AI demand meets physical electricity deployment. If the market keeps rewarding companies that can solve the grid delay problem, $SEI remains one of the more direct tickers to watch.

$WTS — Watts Water Technologies: The Quieter Water And Cooling Infrastructure Angle

Watts’ own corporate overview highlights a disciplined long-term operating profile: $2.4 billion in 2025 net sales, $10.58 in adjusted EPS, 19.6% adjusted operating margin and strong five-year free-cash-flow conversion. Its 2025 annual report frames the company around differentiated water solutions and operating discipline rather than a speculative single-market story. That matters because $WTS gives the article a quality industrial anchor.

The near-term numbers support the quality angle. Watts reported record Q1 2026 results, with net sales of $677 million, up 21% on a reported basis and 12% organically, according to the company’s Q1 2026 earnings materials. Operating margin improved to 19.6%, and diluted EPS increased to $2.97 from $2.21 a year earlier. The company is not being carried only by AI headlines; it is executing across a larger industrial platform.

The AI angle comes from the rising importance of water, liquid cooling and flow-control reliability in dense compute environments. AI servers generate enormous heat. As rack densities rise, traditional air-cooling approaches become less sufficient in many designs, and water-based or liquid-assisted cooling systems become more important. Even when hyperscalers try to minimize direct water consumption, they still need reliable systems for pressure, temperature, flow, backflow prevention, valves, monitoring and water safety across campuses and adjacent facilities.

This makes $WTS a second-order beneficiary rather than a first-order hype name. That can be a strength. The stock does not need every data-center headline to become a company-making event. It can benefit from a broad industrial upgrade cycle while participating in data-center cooling and water efficiency as a growth layer. For readers who are skeptical of overheated AI pure plays, that makes Watts a more balanced way to follow the infrastructure story.

The risk is that the market may overstate the AI exposure. Watts is not a data-center pure play. Its results are influenced by broader construction, commercial, residential, industrial and geographic trends. If an article treats $WTS as if it were a direct fuel-cell or data-center power supplier, that would be misleading. The right framing is more precise: Watts is a high-quality water infrastructure company whose relevance increases as data-center water and cooling become more visible constraints.

That precision matters editorially. The AI power bottleneck is not only about electricity. It is also about the systems that make electricity usable inside a hot, dense, mission-critical facility. Water and cooling do not get the same attention as GPUs, but without thermal management, compute cannot scale. Watts belongs in the article because it represents this quieter but essential part of the infrastructure stack.

$BE — Bloom Energy: The Benchmark For The AI Power Repricing

Bloom’s AI validation is unusually explicit. In April 2026, Bloom and Oracle announced an expanded strategic partnership to deploy up to 2.8 GW of fuel-cell capacity to accelerate AI infrastructure buildout. The company said an initial 1.2 GW was already deploying across Oracle projects in the United States. Bloom stated that Oracle intends to procure up to 2.8 GW of fuel-cell capacity under the expanded deal, with the first 1.2 GW already contracted and deployment underway.

The Oracle deal followed a major October 2025 announcement with Brookfield. Bloom and Brookfield announced a $5 billion strategic AI infrastructure partnership, with Bloom positioned as a preferred onsite power provider for Brookfield’s global AI factories. Whether every investor agrees with the valuation is a separate question; the strategic message is clear. Large infrastructure capital and large cloud customers are willing to explore onsite power as a way to solve AI deployment constraints.

Bloom’s Q1 2026 results then gave the market a financial accelerant. The company reported Q1 revenue of $751.1 million, up 130.4% year over year, and product revenue of $653.3 million, up 208.4% year over year. It also raised full-year 2026 revenue growth guidance, with the midpoint moving to roughly 80% year-over-year growth from a prior midpoint around 60%. Those are not normal industrial growth rates. They explain why $BE has become the poster child for AI power scarcity.

The bull case is that Bloom’s solid oxide fuel-cell systems can provide fast, modular, reliable onsite power in a world where grid access is becoming a competitive advantage. If AI data-center operators need capacity before utilities can deliver large interconnections, fuel cells may have a role alongside gas turbines, batteries, renewables, nuclear discussions and grid upgrades. Bloom’s customer announcements support the idea that the company has moved from clean-tech optionality to strategic infrastructure relevance.

The bear case is valuation and execution. When a stock becomes the flagship of a new theme, the market can price years of perfect deployment before those deployments fully convert into cash generation. Large GW-scale announcements sound enormous, but investors still need to track project economics, gross margins, working capital, service obligations, financing structures, customer concentration and actual cash flow. A fuel-cell company scaling into hyperscale power demand can deliver extraordinary growth, but it can also disappoint if manufacturing, installation, warranty, fuel economics or margins fail to match market expectations.

For this article, Bloom’s role is less about calling it small-cap and more about using it as the market’s reference point. If $BE can carry a premium because investors believe onsite power is central to AI, then investors will naturally search for smaller or less obvious companies exposed to the same bottleneck. That search is part of what makes $SEI and $WTS interesting.

Why The Market Is Looking Beyond Chips

Chip scarcity was the first bottleneck because it was the most visible. GPUs had long lead times, cloud providers fought for allocation, and AI labs competed for compute. But once the semiconductor supply chain starts expanding, the bottleneck does not disappear. It migrates. The next constraint becomes the data-center campus itself: land, permits, substations, power-purchase agreements, cooling systems, water access and grid interconnection queues.

This is why infrastructure names can rally even when they do not manufacture any AI chip. They are solving the next constraint. Investors have seen this pattern before in other cycles. In a gold rush, the most durable winners are often not only the miners, but also the suppliers of picks, shovels, transport, water and energy. In AI, the “picks and shovels” are not just semiconductors. They are power generators, cooling systems, water controls, transformers, switchgear, engineering firms, utility contractors and grid equipment.

The AI capex cycle is also unusually visible because hyperscalers have made it part of their strategic identity. Every major cloud provider wants to prove it can support AI demand. That creates a race for compute capacity. But compute capacity is not simply a purchase order for servers. It is a construction and power-delivery problem. A project with enough GPUs but insufficient electricity cannot monetize those GPUs at scale.

That is why the power bottleneck trade has broader staying power than many smaller AI software narratives. A small software company may announce an AI product and receive a temporary sentiment boost. But a company with real power contracts, real cooling systems, real water infrastructure or real electrical equipment is connected to the physical buildout. The market can still overpay, but the underlying demand is easier to understand.

The Regulatory And Social Backdrop

The AI power story is not frictionless. In fact, one reason it is investable is that it is full of friction. Power projects require permits. Grid upgrades require cost allocation. Local communities worry about electricity bills, water use, air quality and land consumption. Regulators want AI infrastructure to move quickly but also need to protect households and industrial customers from paying the full cost of private data-center expansions.

The June 18 FERC development is important because it shows the federal regulatory system is now engaging directly with the large-load connection problem. Grid operators must address how very large customers connect and how infrastructure costs are allocated. That could accelerate some projects, but it could also create clearer obligations for data-center customers. From an investment standpoint, better rules may help legitimate infrastructure providers, while exposing weaker projects that rely on opacity or unrealistic assumptions.

The Reuters investigation into fast-tracked off-grid plants adds another layer. If local opposition builds, some projects may slow. If environmental scrutiny increases, fuel choice and emissions profiles may become more important. If communities push back against water usage, cooling and water-efficiency technologies may gain value. This is where the trade can broaden from simple “more power” to “better, faster, more acceptable infrastructure.”

That distinction matters. The winners may not be the companies that provide the cheapest megawatt at any cost. They may be the companies that provide reliable power, acceptable permitting profiles, credible emissions management, efficient water use, rapid deployment and transparent project structures. Solaris, Watts and Bloom each touch a different part of that challenge.

Competitive Landscape

The AI power bottleneck basket is much larger than $SEI, $WTS and $BE. Those three are the clean Stocktwits cashtags for this article, but the broader map includes utilities, power producers, electrical equipment companies, engineering and construction firms, cooling specialists and energy storage companies.

On the power generation side, large players such as GE Vernova, Caterpillar, Cummins and Siemens Energy are relevant because turbines, generators and grid equipment are becoming central to data-center deployment. On the electrical infrastructure side, Eaton, Vertiv, Schneider Electric and Quanta Services sit closer to power distribution, switchgear, UPS, thermal management, engineering and grid buildout. On the power supplier side, utilities and independent power producers such as Constellation, Vistra, Talen and others remain important because hyperscalers increasingly need large blocks of firm power.

Bloom competes against these broader alternatives. Fuel cells are only one solution. Gas turbines may be cheaper and more familiar. Batteries can support resilience but are not always a full generation substitute. Nuclear power is strategically attractive but slower in many real-world deployments. Renewables are essential but require storage, firming and grid integration. The AI power stack will probably be hybrid rather than dominated by a single technology.

Solaris also faces competition. Behind-the-meter power is attractive, but customers will compare pricing, reliability, emissions, speed, financing and technical risk. If larger industrial companies decide to attack the same opportunity aggressively, Solaris will need to prove that its customer relationships, execution speed and turnkey model create a durable edge.

Watts competes in a more mature industrial context. Its advantage is not that it owns the AI cooling market, but that it has established brands, product breadth, distribution and operating discipline. The company may benefit from the cooling and water-efficiency cycle, but investors should not expect it to behave like a high-beta AI pure play every day.

Catalysts To Watch

The first catalyst is contract conversion. For Solaris, investors should watch whether announced long-term power contracts begin on schedule, scale as expected and translate into reported revenue, EBITDA and free cash flow. The >600 MW contract beginning in late 2026 and scaling through 2028 is particularly important because it can help validate the market’s confidence in the company’s execution model.

The second catalyst is guidance. Companies in this theme are being valued partly on future capacity and future demand. If management teams raise revenue, EBITDA or margin guidance, the market may continue to reward the group. If guidance becomes more cautious, especially after large stock moves, investors may quickly question valuation.

The third catalyst is policy. FERC’s large-load connection review, state-level data-center permitting policies, utility cost-allocation decisions and environmental reviews can all influence deployment speed. A policy framework that clarifies connection rules may support infrastructure providers, but new restrictions or community pushback could slow specific projects.

The fourth catalyst is hyperscaler capex commentary. If Oracle, Microsoft, Amazon, Google, Meta or other major data-center builders continue to signal heavy AI infrastructure spending, the power bottleneck trade stays alive. If hyperscalers suddenly pull back or delay capex, second-order infrastructure names could derate.

The fifth catalyst is water and cooling visibility. Any major hyperscaler announcement around water-positive goals, liquid cooling adoption, closed-loop systems, water recycling, heat reuse or thermal infrastructure could support names linked to water and cooling systems. That is especially relevant for $WTS because its AI angle depends on investors caring about the water/cooling layer, not just generation.

Bull Case

In the bull case, the market continues to broaden the AI trade away from semiconductors and toward physical infrastructure. This would favor companies with tangible products and services that solve deployment constraints. Solaris could benefit from more long-term power contracts with investment-grade technology customers. Watts could benefit from rising data-center cooling and water-management needs layered on top of its existing industrial platform. Bloom could benefit from Oracle, Brookfield and other customers validating the use of fuel cells as a major onsite power solution.

The strongest version of the bull case is not just earnings growth. It is multiple expansion. Investors may decide that certain industrial infrastructure companies deserve higher valuations because they are attached to AI capex rather than only traditional construction or energy cycles. That has already happened in parts of the electrical equipment and power-generation universe. If it continues, the group can remain in focus even after large moves.

Bear Case And Red Flags

The most obvious red flag is valuation. A real theme can still produce bad entries. Bloom is the clearest example: the business has powerful strategic validation, but the stock has already been repriced dramatically. When a company becomes the flagship of a hot narrative, expectations can become unforgiving. Even good results may not be enough if the market has already priced perfection.

Solaris has a different risk profile. It is smaller and more direct, which makes it more exciting, but also more exposed to execution and capital structure. Scaling generation capacity requires equipment, financing, logistics, customer coordination and technical delivery. If costs rise or timelines slip, the same operating leverage that excites investors can work in reverse.

Watts has the opposite problem. It is high quality but less direct. If the market rotates into pure power names, $WTS may lag because investors view it as too diversified. If construction or industrial demand softens, AI-related cooling exposure may not be enough to offset broader pressure. That does not make Watts weak; it just means investors should understand what they own.

Another red flag is policy backlash. The power bottleneck trade benefits from urgency, but urgency can produce sloppy permitting and local resentment. If communities feel that data centers are receiving special treatment while residents face higher bills, water stress or pollution risk, political pressure can rise. This may create opportunities for cleaner, more efficient technologies, but it can also delay projects across the board.

Merlintrader Bottom Line

The AI power bottleneck trade is one of the more credible second-order AI themes in the market because it is grounded in a real physical constraint. This is not a vague “AI transformation” story. It is a concrete deployment problem: data centers need power, cooling, water systems and grid access faster than the traditional infrastructure cycle can comfortably provide.

$SEI is the most direct small/mid-cap expression of the theme, with behind-the-meter power contracts that make the AI infrastructure story tangible. $WTS is the quieter quality-industrial angle, tied to water and flow-control systems that matter more as cooling and water efficiency become central to data-center development. $BE is the liquid benchmark and headline leader, but its current scale and valuation mean it should be treated differently from a classic emerging small/mid-cap idea.

The clean editorial framing is this: AI is not only a semiconductor trade anymore. It is becoming an electricity, water and infrastructure trade. The most interesting opportunity may come from identifying which companies are solving actual bottlenecks rather than simply borrowing the AI label. Solaris, Watts and Bloom each deserve attention for that reason, but each also carries a different kind of risk.

For Stocktwits distribution, the three cashtags are clean and coherent: $SEI $WTS $BE. Inside the article, the broader watchlist can include power equipment, utilities, cooling, fuel-cell, generator and grid names, but the public-facing ticker discipline should stay tight. Three cashtags, one clear theme, no confusion.