AI data centers are running out of electricity from the traditional power grid, forcing tech giants to build their own private power plants using natural gas turbines. SpaceX's xAI division recently disclosed plans to spend more than $2.8 billion on natural gas turbines over the next three years, a move that underscores an emerging trend in AI infrastructure: hyperscalers are increasingly turning to on-site, behind-the-meter energy generation to power their massive computing operations.

This shift reflects what some analysts call the "Dark Energy" thesis, a concept that describes the use of natural gas turbines as a solution to AI's growing power demands. These turbines function like jet engines adapted into electric generators, can start in minutes, and produce tens of megawatts of power per unit. Unlike waiting years for grid connections, companies can deploy these systems quickly to power their data centers independently.

Why Is the Power Grid Failing AI Data Centers?

The U.S. power grid simply cannot keep up with the electricity demands of modern AI infrastructure. Utilities in major cities across the nation are already rationing power, and some municipalities have imposed moratoriums on new data center projects until 2030 or beyond. Tech companies racing to train ever-larger AI models face a stark choice: wait upward of five years for a grid connection, or bring a private power plant directly to the data center.

SpaceX's filing reveals just how critical this bottleneck has become. The company stated in its initial public offering documents: "We currently rely significantly on natural gas and gas turbine technology to power our data center operations." The filing also warned that any injunction or revoked permit could damage its AI business, highlighting the regulatory risks these companies face.

What Are the Key Drivers Behind This Infrastructure Shift?

The scale of investment in AI infrastructure is staggering. Global hyperscale data center investments are projected to exceed $1.2 trillion between 2025 and 2027, creating unprecedented demand for power generation and cooling systems. This massive capital expenditure cycle is being compared to historical infrastructure booms like the global 4G wireless rollout, the post-2008 liquefied natural gas build-out, and the early-2010s shale boom.

Beyond power generation, the AI infrastructure buildout is creating opportunities across an entire supply chain. Companies that manufacture optical networking equipment, cooling systems, transformers, switchgear, and power components are experiencing explosive growth. In India alone, an equal-weighted index of 28 companies feeding the data center ecosystem has added approximately $47 billion in combined market value this year, a rise of nearly 50 percent.

How to Understand the Hidden Winners in AI Infrastructure?

• Optical Networking Suppliers: Companies like Applied Optoelectronics manufacture high-speed optical transceivers and connectivity solutions that move massive amounts of data between graphics processing units (GPUs), servers, and storage systems at extremely high speeds. Demand for 400G and 800G optical modules is expected to exceed production capacity through mid-2027.
• Power and Cooling Component Makers: Manufacturers of precision cooling systems, transformers, switchgear, and related equipment are experiencing multi-year backlogs. These components have two-to-four year lead times, creating what analysts call an "enviable seller's market" with orders secured now bringing revenue between 2027 and 2029.
• Fiber Optic and Cable Manufacturers: Sterlite Technologies, an optical fiber maker, surged more than 530 percent this year after securing a $1.1 billion multi-year contract from a U.S. hyperscaler. Its competitor HFCL has jumped 191 percent, demonstrating the strength of demand across the supply chain.

The market is rewarding companies with visible AI-linked earnings rather than just thematic exposure. However, valuations have become stretched in some cases. Sterlite is trading at approximately 70 times its 12-month forward earnings, compared to the broader market's 19 times, leaving little room for execution disappointments.

What Legal and Environmental Challenges Are Emerging?

Musk's aggressive push toward natural gas turbines has already drawn legal scrutiny. The NAACP, joined by local environmental groups, sued xAI over its operation of gas turbines at AI data centers. The Southern Environmental Law Center argues that the "mobile" turbines are being operated in violation of federal law, which states that power plants mounted on a trailer can still be considered stationary and subject to air pollution regulations.

So far, xAI has been granted permits for just 15 turbines while operating up to 46, creating a significant regulatory gap. This legal uncertainty hasn't deterred Musk's investment, however, suggesting that the power crisis has become desperate enough to justify the regulatory risk.

Beyond individual lawsuits, a broader backlash against data center construction is gaining momentum. According to Data Center Watch, a research firm tracking opposition, at least 142 activist groups across 24 states are organizing to block data center construction. Over the past two years, residents have blocked or delayed approximately $64 billion worth of data center projects. Notably, roughly 55 percent of elected officials who have spoken out against data centers are Republicans, indicating that opposition has become bipartisan rather than following traditional political fault lines.

Why Are Suppliers More Attractive Than Hyperscalers?

Analysts increasingly suggest that the biggest investing opportunities lie not with the headline-grabbing hyperscalers themselves, but with the suppliers that provide the critical components and infrastructure. When a new technology creates bottlenecks, the companies that solve those bottlenecks often experience outsized gains.

"We may be on the wrong end of the AI trade, but we could be on the right side of the AI capex trade. One could consider companies benefiting from data centers and the entire value chain associated with this capex," said R. Sivakumar, chief investment officer at Axis Mutual Fund.

R. Sivakumar, Chief Investment Officer at Axis Mutual Fund

Foreign investors are already recognizing this opportunity. Shareholding of foreign funds in Indian industrial stocks rose to 14 percent as of end-March, the highest in two years, even as global funds remain record sellers of Indian stocks overall. This divergence suggests sophisticated investors are specifically targeting the AI infrastructure supply chain.

The data center capex cycle is expected to remain robust for years. As enterprises and cloud providers deploy larger AI models and inference workloads, they must build more interconnected data center capacity, increasing the need for optical networking solutions, power generation equipment, and cooling systems throughout the ecosystem. If AI-related capital expenditures remain strong over the coming years, suppliers could see sustained demand growth and improved revenue visibility.

Musk's $2.8 billion turbine purchase is ultimately a confirmation that the power bottleneck has become the defining constraint on AI infrastructure expansion. The companies that solve this constraint, whether through turbine manufacturing, optical networking, cooling systems, or power distribution equipment, are positioned to capture significant value from the largest industrial investment cycle in decades.