Will AI and Data Centers Drive Investor Returns?

The question isn't whether AI is reshaping data center economics; it already has. The real question investors need to answer is whether the companies positioned at the intersection of artificial intelligence infrastructure and power delivery can sustain the margin expansion that has made them Wall Street darlings — or whether the capital intensity of this buildout will eventually eat what it created.

GE Vernova's recent acquisition activity signals something worth paying attention to. When an industrial giant makes strategic moves during a period of explosive AI infrastructure demand, it's rarely coincidental. Investors are watching closely to see whether AI and data centers continue to fuel orders and margins — and what the answers mean for anyone with capital allocated to this space.

The Role of AI in Data Center Operations

Modern data centers are no longer passive warehouses for servers. They are active, dynamic compute environments where AI workloads — training large language models, running inference at scale, processing real-time analytics — demand fundamentally different infrastructure than the enterprise IT loads of a decade ago.

The shift from general-purpose computing to AI-optimized workloads isn't a software problem; it's a power problem.

A traditional enterprise server rack might draw 5 to 10 kilowatts. An AI training cluster using NVIDIA H100 GPUs can push 40 to 80 kilowatts per rack — sometimes more. Multiply that across a hyperscale campus, and you're talking about facilities that consume as much electricity as a small city. Microsoft, Google, Amazon, and Meta are collectively spending hundreds of billions of dollars on data center buildout through 2026 and beyond, and a significant portion of that spend flows directly into power infrastructure.

This is where the operational efficiency argument gets interesting. AI isn't just driving demand for data centers — it's also being deployed *within* data centers to optimize cooling, predict hardware failures, and manage power draw dynamically. The same technology creating the demand is being used to make the infrastructure more sustainable. That dual role matters for long-term data center profitability because energy costs typically represent 40 to 60 percent of operating expenses for a hyperscale facility.

Current Trends in Data Center Investments

The numbers are staggering, and context makes them even more so. Global data center investment surpassed $300 billion annually in recent years, and AI-driven demand is accelerating, not plateauing. Northern Virginia — "Data Center Alley" — has seen land prices and power availability become genuine constraints, pushing development into secondary markets like the Carolinas, Texas, Ohio, and the Pacific Northwest.

What's less obvious is that the bottleneck has shifted from capital to kilowatts.

Developers can raise the equity. They can secure the debt. What they can't always secure is a utility interconnection agreement on a timeline that matches their construction schedule. Grid interconnection queues in the U.S. have ballooned to over 2,700 gigawatts of pending requests — a figure that dwarfs the existing installed generation capacity of roughly 1,200 gigawatts. Data center developers competing for grid access are, in a real sense, competing against solar farms, wind projects, and battery storage assets for the same limited interconnection slots.

This dynamic is precisely why companies like GE Vernova — which sits at the nexus of power generation equipment, grid infrastructure, and electrification technology — are attracting so much investor attention. Vernova's acquisition strategy reflects a calculated bet that the companies controlling the power delivery stack will capture durable value as AI infrastructure spending scales. The key players aren't just the hyperscalers building campuses; they're the turbine manufacturers, transformer suppliers, switchgear companies, and grid software providers enabling those campuses to actually turn on.

Financial Implications of AI for Investors

From a pure AI market impact standpoint, the investment thesis is compelling on the surface: surging demand, limited supply of power-ready land, and long-term contracted revenue from creditworthy tenants. But the financial picture has meaningful nuance.

Margin expansion in data center-adjacent businesses — power equipment, cooling systems, backup generation — has been real and significant. Order backlogs at industrial suppliers have stretched to two, three, even four years in some categories. That kind of visibility is unusual in capital goods manufacturing and gives investors unusual confidence in near-term revenue.

The risk isn't demand; the risk is whether the supply chain can scale fast enough to avoid margin-eroding bottlenecks.

Transformer shortages are the canonical example. U.S. transformer manufacturers are booked out years in advance, creating delays that can push data center commissioning timelines by 12 to 18 months. Those delays don't kill projects — the demand is real — but they create cash flow timing issues that investors in development-stage assets need to price correctly.

For publicly traded infrastructure and industrial companies levered to AI data center growth, the long-term forecast is broadly positive. Power demand in the U.S. was essentially flat for two decades. That era is over. Goldman Sachs estimated that data centers could account for 8 percent of U.S. electricity consumption by 2030, up from roughly 3 percent today. That's a structural shift that plays out over years, not quarters — which means the investment horizon matters enormously in how you evaluate any given position.

Challenges Facing Data Centers with AI Adoption

The infrastructure demands of AI-scale compute are creating friction at every layer of the stack, and not all of it is solvable with capital.

Water consumption is a real issue that's becoming a regulatory flashpoint. Hyperscale data centers using evaporative cooling can consume millions of gallons per day, and municipalities in water-stressed regions are increasingly pushing back. Arizona has seen direct conflict between data center developers and local water authorities. This isn't an abstract ESG concern — it's a permitting and site selection constraint that materially affects where the next generation of AI infrastructure gets built.

The regulatory environment around energy use is also evolving. The EU's Energy Efficiency Directive has placed new reporting and efficiency requirements on large data centers. U.S. federal policy is less prescriptive, but state-level action is accelerating, particularly in markets with stressed grids. Developers who treated regulatory risk as a checkbox item are learning it deserves a seat at the underwriting table.

On the grid side, the push toward on-site or behind-the-meter power generation — small modular reactors, natural gas turbines, fuel cells — reflects the industry's acknowledgment that utility interconnection timelines are a genuine constraint. Microsoft's deal with Constellation to restart Three Mile Island and Amazon's investments in SMR developers aren't PR stunts; they're infrastructure strategies driven by grid access scarcity.

What Investors Should Actually Watch

Forecasting the evolution of AI infrastructure investment is less about predicting technology adoption curves — that adoption is happening faster than most projections — and more about identifying where value accrues in the supply chain.

The companies that will outperform are those controlling scarce inputs: permitted land with power, long-lead equipment manufacturing capacity, and grid interconnection rights.

For investors with exposure to this theme, a few specific things warrant close attention. First, watch order backlogs and lead times at power equipment manufacturers. Expanding backlogs at healthy margins signal genuine supply constraints; contracting backlogs signal the market is finding equilibrium or that a specific supplier is losing share. Second, track interconnection queue reform. The Federal Energy Regulatory Commission's Order 2023 attempted to streamline the interconnection process — its real-world implementation will determine whether the grid bottleneck eases or persists. Third, monitor hyperscaler capex guidance. When Microsoft, Google, or Amazon revise their data center spending commitments up or down, the signal travels quickly through the entire supply chain.

GE Vernova's strategic positioning — and its acquisition activity — reflects an industrial company that has read this landscape correctly. The AI and data center investment thesis isn't speculative anymore; it's infrastructure. The question for investors isn't whether to have exposure, but how to size it, where in the value chain to own it, and how to think about the capital cycle risks that come with any major buildout.

The buildout is real. The opportunity is real. The discipline required to invest in it well is where most people will be tested.

Explore more about investment opportunities in AI and data centers here.

INTERNAL LINK SUGGESTIONS

• [INTERNAL LINK: AI infrastructure trends]
• [INTERNAL LINK: data center investment strategies]
• [INTERNAL LINK: power delivery innovations]