Identifying the Infrastructure Layer Stocks That Will Define Tech by 2028

The AI revolution is hitting a fundamental S-curve inflection point. While the software and chip layers capture headlines, the exponential growth of compute is now being throttled by a physical bottleneck: power. The infrastructure layer is becoming the primary constraint on adoption, setting the stage for a massive, multi-year build-out that will define winners.

The numbers reveal a staggering trajectory. Global AI data center power demand is projected to reach 68 GW by 2027, nearly doubling the total data center power from 2022. This isn't a gradual climb; it's an exponential surge. In the United States, the strain is already visible. The country faces a projected 36 GW power shortfall by 2028. Real-world grid instability is the early warning sign. In July 2024, a voltage fluctuation in northern Virginia triggered the simultaneous disconnection of 60 data centers, forcing emergency adjustments to prevent cascading outages. This is the physical layer cracking under the weight of the digital boom.

This pressure is fueling an unprecedented capital rush. Hyperscalers are spending $100–$200 billion in just six months on AI hardware and data centers alone. The scale of this build-out is historic, often likened to the most significant industrial booms of the past. Yet, even this massive spending is racing against a fixed physical reality. The investment thesis here is clear: the AI energy crunch is a multi-year infrastructure opportunity. Companies that provide the fundamental rails-new power generation, transmission, and especially energy storage-will be the picks and shovels of the AI era. The bottleneck is real, and the build-out has only just begun.

The Infrastructure Layer: Solving the Energy Crunch

The AI energy bottleneck is forcing a radical rethinking of power infrastructure. The solution isn't just building more power plants; it's about creating a new, resilient layer of energy systems that can deliver the pristine, always-on electricity AI servers demand. This is where the true picks and shovels are emerging.

Energy storage is the critical, hidden growth engine. It forms a $50-billion total addressable market by 2030 specifically to provide reliable, jitter-free power for data centers. Companies like Fluence Energy are positioned to capitalize, with its industrial-scale battery modules designed to smooth out renewable energy and ensure data center stability. The opportunity is massive, with the company already developing 36 gigawatt-hours of data center projects. This isn't just backup power; it's the foundational layer for a new grid.

On-site power solutions are booming as a quick fix for the gridlock. The "Bring Your Own Power" boom is a direct response to permitting snarls and a grid that can't keep pace. In West Texas, natural-gas-fired power generation is being built for OpenAI's Stargate project. More than a dozen Equinix data centers are using fuel cells for power. The scale of this DIY energy rush is staggering, with Bloom Energy securing a $5 billion partnership to be the preferred onsite power provider for a major AI data center buildout. This is the infrastructure layer in action, bypassing the slow-moving utility grid entirely.

Then there's a new frontier entirely. Google's Project Suncatcher explores a moonshot: equipping solar-powered satellite constellations with TPUs to scale machine learning compute beyond Earth's limits. The premise is simple-space offers nearly continuous, high-productivity solar energy. This isn't science fiction for 2050; it's early research into a potential future infrastructure layer for the most extreme compute demands. It represents the ultimate paradigm shift in where we source the energy for the next paradigm.

The bottom line is that the AI energy crunch is a multi-year infrastructure build-out. The winners will be the companies providing the fundamental rails: energy storage, on-site generation, and eventually, even space-based power. The physical layer is the new bottleneck, and solving it defines the next phase of the technological S-curve.

Candidate Stocks: The Infrastructure Layer Picks

The structural build-out for AI energy infrastructure is creating a new class of picks and shovels. These are not speculative bets on future tech; they are concrete plays on the physical rails that must be laid to fuel the next paradigm. Three companies stand out as direct beneficiaries of this multi-year boom.

First is New Era Energy & Digital (IREN). The company is a pure-play enabler of hyperscaler power demand. Its investment case is anchored by a massive, contracted pipeline. IREN secured a $1.94 billion in annual recurring revenue through a 5-year deal with Microsoft for 200 megawatts. More importantly, it has a 4.5 gigawatt pipeline after acquiring 1.6 gigawatts of power in Oklahoma. That pipeline represents a potential revenue stream of over $40 billion annually, dwarfing its current market cap. The company is monetizing the grid shortfall directly, turning a fixed physical constraint into a scalable revenue engine.

Next is Bloom Energy (BE), a key player in the "Bring Your Own Power" boom. As utility grids struggle to keep pace, companies are building on-site generation to bypass permitting delays. Bloom Energy is positioned to be the preferred provider for a major AI data center buildout, having secured a $5 billion partnership. Its fuel cell technology offers a reliable, cleaner alternative to diesel generators for data center backup. This partnership is a direct bet on the DIY energy rush, capturing value from the rapid deployment of power solutions that cannot wait for the slow-moving grid.

Finally, Fluence Energy (FLNC) addresses the critical need for pristine, always-on power. Its industrial-scale battery modules are essential for smoothing renewable energy and ensuring data center stability. The company is already developing 36 gigawatt-hours of data center projects. While its official order backlog may not yet reflect this, the CEO has identified data centers as the company's biggest new customer segment. Fluence is the hidden growth engine, providing the energy storage layer that makes the AI power supply chain reliable and resilient.

These three stocks represent different but complementary points on the infrastructure S-curve. IREN provides the foundational power supply, Bloom Energy offers the on-site generation bypass, and Fluence ensures the quality and continuity of that power. Together, they are building the physical layer that will define the technological landscape for years to come.

Financial Impact, Catalysts, and Risks

The structural shift from AI chips to underlying infrastructure is a multi-year capital reallocation. The total investment in AI-centric chips, servers, and facilities could approach $3 trillion by 2028. This isn't a one-time build-out; it's a sustained boom that will redirect capital from the semiconductor layer to the physical rails that power it. For infrastructure providers, this means a massive, multi-decade revenue opportunity.

The near-term catalyst that will confirm the thesis is the signing of major long-term power purchase agreements (PPAs) by hyperscalers. These contracts de-risk the massive build-out for infrastructure companies. When a giant like Microsoft or Google locks in a 10- or 20-year power supply deal, it provides the predictable revenue stream needed to finance new power plants, transmission lines, and storage facilities. The market is watching for these milestones as proof that the demand surge is real and durable, not a speculative bubble.

Yet the path is fraught with friction. The primary risk is permitting. Getting a new power line or a grid connection approved can take four years or more. This regulatory bottleneck is a major source of delay and cost, forcing companies to seek alternative solutions. This is fueling the "energy Wild West" where tech firms are building their own power plants and fuel cells to bypass the slow-moving grid. While this DIY rush creates immediate demand for companies like Bloom Energy, it also introduces volatility and potential for stranded assets if regulatory overhangs eventually force a return to centralized planning.

Another key risk is the bullwhip effect in reverse. The current frenzy of capital expenditure could lead to overcapacity if demand growth slows. Data center electricity consumption is projected to grow from 176 terawatt hours in 2023 to between 325-580 TWh by 2028. If the upper end of that range proves too optimistic, utilities and infrastructure providers could face stranded costs. The early signs of grid instability-like the simultaneous disconnection of 60 data centers in July 2024-show the system is already under stress. The investment case hinges on the demand trajectory staying on the high end of that range.

The bottom line is that the financial opportunity is enormous, but the execution is messy. The catalysts are clear: long-term PPAs and the continued scaling of the "Bring Your Own Power" model. The risks are equally clear: permitting delays, regulatory shifts, and the specter of overcapacity. For investors, this is a high-stakes bet on the physical layer of the next technological paradigm. Success requires navigating a multi-year build-out where the rails are being laid as the train is already moving.