Major technology companies that championed renewable energy are now locking in decades-long contracts for natural gas and nuclear power to support their massive artificial intelligence data centers. Microsoft signed a 20-year agreement with Chevron for a 2.7-gigawatt natural gas complex in Texas, while Meta is funding roughly 10 new natural gas plants totaling 7.5 gigawatts in Louisiana. Both companies also committed to nuclear energy, with Microsoft restarting the Three Mile Island reactor in Pennsylvania and Meta securing a 20-year deal for the Clinton nuclear plant in Illinois.

The shift reveals a fundamental tension in the energy transition: artificial intelligence requires continuous, reliable power 24 hours a day, 365 days a year. Wind and solar, while increasingly cost-effective, cannot guarantee electricity on demand when the weather does not cooperate. For companies investing hundreds of billions of dollars in server infrastructure, that uncertainty is unacceptable.

Why Are Tech Giants Abandoning Their Green Energy Commitments?

For years, companies like Microsoft and Meta were vocal advocates for renewable energy. Microsoft pledged to match its electricity use with 100 percent renewable energy and become carbon negative by 2030. Meta joined RE100, a global corporate initiative whose members commit to 100 percent renewable electricity. Both companies poured billions into renewable projects and used their political influence to support climate policies.

Then artificial intelligence arrived at scale. The computational demands of training and running large language models (LLMs), which are AI systems that process and generate human language, require staggering amounts of electricity. A single data center can consume as much power as a small city. When faced with the operational reality of keeping servers running continuously, corporate sustainability commitments collided with engineering necessity.

Meta quit RE100 last month after roughly a decade of membership. The company still claims it matches its electricity use with clean and renewable energy, but largely through certificates and contracts rather than actual power generation. As one energy analyst noted, matching renewable energy on paper over a year is fundamentally different from generating power at 2 a.m. on a windless night in Louisiana.

What Does This Mean for Nuclear and Natural Gas Markets?

The energy commitments from Microsoft, Meta, and other hyperscalers (large-scale cloud computing companies) are reshaping investment expectations for both nuclear and natural gas sectors. Elon Musk's recent comments about natural gas signal how even the most ambitious clean energy advocates are embracing a more pragmatic approach.

Musk, who built Tesla's mission around accelerating the death of the "mine-and-burn hydrocarbon economy," acknowledged in late July that natural gas may serve as a long-term bridge fuel, particularly to offset seasonal solar declines during winter months. SpaceX is building a dedicated natural gas pipeline for rocket launches while scaling AI compute infrastructure that demands enormous power. During SpaceX's first investor call in August, Musk revealed the company is targeting 15 to 20 gigawatts of power capacity by the end of next year, exceeding New York City's peak demand of roughly 10 gigawatts.

This creates a complex investment landscape for nuclear developers like Oklo and NuScale Power, which have benefited from AI-driven enthusiasm for small modular reactors (SMRs). Bank of America analysts frame nuclear energy as a $10 trillion global opportunity over coming decades, driven by AI's need for massive amounts of clean, baseload power. However, traditional nuclear plants often take a decade or more to complete, leaving a significant timing gap where natural gas can fill immediate demand.

How Are Data Centers Building Their Own Power Infrastructure?

One of the most striking developments is that major technology companies are no longer relying solely on the electrical grid. Instead, they are building their own dedicated power generation on-site, a trend that has profound implications for utilities and ratepayers.

• On-Site Generation Scale: Approximately 25 percent of data center capacity under development plans to build its own generation infrastructure, representing roughly 90 gigawatts across 59 tracked projects.
• Grid Independence Strategy: Big Tech is not just buying reliable power; it is buying its way off a grid that its own lobbying helped make less reliable and more expensive through support for weather-dependent renewable generation.
• Ratepayer Impact: Families, factories, hospitals, and farms cannot build their own power plants and remain stuck with the system that policy created, one that piled on weather-dependent generation while retiring coal and nuclear plants that ran on demand.

This divergence between what large corporations can afford and what ordinary consumers can access raises a critical question about energy equity. Ratepayers are paying for two systems (reliable baseload and renewable generation) but getting one, while their electricity bills keep climbing. Utilities are asking regulators to approve billions in new plants and transmission wires to support data center development, but who bears the cost if that demand falls short ?

What Does the Energy Transition Actually Look Like Now?

The emerging picture is one of a layered, portfolio-based energy strategy rather than a binary choice between fossil fuels and clean power. SpaceX's approach illustrates this complexity: the company relies on a diverse energy mix including wind, solar, and natural gas to support its data center buildout. Other technology giants have been more aggressive on nuclear, but the pattern is consistent across the industry.

Bank of America's analysis suggests that the sheer magnitude of expected demand growth means there may be room for multiple power sources to scale simultaneously. If natural gas remains a durable part of the energy mix, the urgency to deploy nuclear capacity could diminish at the margins, potentially weighing on valuations for pre-revenue SMR developers that trade on future growth expectations. Conversely, the competitive pressure created by Musk's comments and other corporate power announcements could accelerate procurement decisions across the tech sector.

The broader implication is that electricity is only valuable when it is there the moment you need it. For decades, energy policy focused on the source of power (coal, natural gas, renewables, nuclear) rather than its reliability. AI data centers, electric vehicles, industrial electrification, and space infrastructure are all competing for finite power resources, and companies with the capital to do so are solving the problem themselves by building dedicated generation capacity.

Whether nuclear or natural gas captures the largest incremental share of new demand will depend on execution speed, cost curves, and regulatory environments, all of which remain highly uncertain. What is clear is that the energy transition is no longer a simple story of replacing fossil fuels with renewables. It is a complex, pragmatic calculation where reliability, speed, and cost matter as much as carbon emissions.