Are AI Companies Building Their Own Power Plants? Gigawatt Plans

Are AI Companies Building Their Own Power Plants? Gigawatt Plans

The global race for Artificial Intelligence dominance has quietly triggered an unprecedented energy land grab, pushing tech giants to bypass traditional utility grids and invest directly in their own power infrastructure. This isn't just about more electricity; it's about the specific, continuous, and immense power demands of AI training and inference, which are driving a new wave of localized, advanced energy solutions, from dedicated renewables to small modular reactors (SMRs). What I’ve found in my research is a dramatic and accelerating shift in how the world’s most powerful companies are securing their future.

The Unprecedented AI Power Spike

By 2025, AI-focused data centers consumed an estimated 155 terawatt-hours (TWh) of electricity, representing about 0.5% of the world's total electricity. This figure is projected to skyrocket, with data centers — largely driven by AI — expected to consume up to 3% of global electricity by 2030, reaching approximately 945 TWh. In some regions, like the U.S., I found that data centers could account for 9% to 17% of total electricity demand by 2030, up from roughly 3-4% today. More specifically, U.S. data centers consumed 176 TWh in 2023, making up about 4.4% of total U.S. electricity use. This is projected to grow to between 325 and 580 TWh, or 6.7% to 12.0% of national electricity consumption, by 2028. Some estimates even suggest U.S. data center electricity use could reach 400-600 TWh by 2030.

This growth isn't linear. I discovered that training a single large language model like GPT-4 can consume around 50 GWh. [original, cite: 26] While inference (the running of AI models) is becoming the dominant driver of energy usage, accounting for 70-90% of lifecycle energy as AI becomes ubiquitous, it also introduces highly volatile and spiky load profiles. AI workloads can cause power demand swings of 40-50% over short periods, a challenge traditional grid infrastructure is ill-equipped to handle. [original] The power capacity of leading AI supercomputers has doubled every 13 months, with xAI's Colossus supercomputer alone using 280 MW, often relying on mobile generators due to insufficient local grid capacity. [original] A single modern AI data center can demand as much power as 100,000 homes, and I've seen that some larger facilities currently under construction are expected to consume up to 20 times that amount. In fact, Goldman Sachs Research forecasts global power demand from data centers to rise by a staggering 165% by 2030 from 2023 levels.

Big Tech's Bold Energy Pivot: Beyond the Grid

Recognizing that traditional grid expansion simply cannot keep pace (wait times for grid interconnections have escalated to 10 years or more in some regions), major tech companies are fundamentally reshaping their energy strategies. [original] I found that projects entering service in 2025 took an average of more than seven years to reach operational status, according to PJM Interconnection data. The bottleneck, I’ve learned, has shifted downstream from interconnection queues to transmission buildouts, substation capacity, and strained supply chains. For example, substation transformer lead times averaged 140 weeks in 2023, increased to about 150 weeks in 2025, and now exceed 160 weeks in 2026. Nearly 2,300 GW of generation and storage capacity were stuck in U.S. interconnection queues at the end of 2024, which is more than the country's entire existing generating fleet. Analysts like Gartner predict that power shortages will restrict 40% of AI data centers by 2027.

This critical situation has pushed companies to make massive investments. Microsoft, for example, has pivoted from passive energy procurement to direct investment in large-scale power infrastructure, earmarking $80 billion for AI-enabled data centers in fiscal year 2025, treating energy access as a competitive advantage. [original, cite: 35] I also discovered that Microsoft has committed $80 billion through 2028 to expand AI-optimized data centers and Azure cloud regions worldwide, which is its largest infrastructure investment ever. This includes direct partnerships with energy producers and financiers to build dedicated power systems. Microsoft is also making significant international investments, including $19 billion CAD in Canada between 2023 and 2027, with more than $7.5 billion CAD in the next two years for new digital and AI infrastructure. In Asia, I found that Microsoft is making its largest investment, committing $17.5 billion over four years (CY 2026 to 2029) in India, on top of an earlier $3 billion investment, with a new hyperscale data center going live in mid-2026.

Amazon, a long-time leader in corporate renewable energy procurement, has also acknowledged that generative AI's increasing demand will necessitate a more aggressive energy strategy. I learned that Amazon has invested in over 700 carbon-free energy projects globally, representing more than 40 GW of capacity, enough to power the equivalent of over 12 million homes in America. In 2025 alone, Amazon contracted 10.22 GW of clean energy. Recently, I found that Amazon announced investments in 700 MW of new carbon-free energy projects in Nevada to power future data center operations, including 100 MW of geothermal power and 600 MW of solar plus 600 MW of battery storage.

Google (Alphabet) is equally aggressive. My research shows that Alphabet invested roughly $90 billion in capital expenditures in 2025 and plans to nearly double that to as much as $185 billion in 2026, with most of that money flowing to data centers and their power infrastructure. In 2024, Google's data centers used 30.8 million MWh of electricity, more than some entire countries. I found a notable example in Texas, where Google plans to invest $40 billion in data centers. This includes a partnership with Crusoe Energy to develop a campus near Claude, known as 'Goodnight,' where Crusoe filed a permit in January 2026 to build a 933 MW natural gas power plant on-site, not connected to the grid. Google is also partnering with AES in Texas, co-locating an 850 MW data center with 600 MW of solar and 945 MW of wind, bypassing traditional grid queues. Furthermore, I’ve seen Google paying $3 billion to keep aging hydroelectric dams on the grid in Pennsylvania and restarting Iowa's only nuclear plant under a 25-year Power Purchase Agreement (PPA).

These hyperscalers—Alphabet, Amazon, Microsoft, and Meta—collectively plan to invest over $350 billion in data centers in 2025 and about $400 billion in 2026. This level of investment highlights a profound shift. I believe the industry is moving towards a model where power generation is becoming an integral part of data center development, often "behind the meter." Natural gas currently leads as the most deployed energy source for new AI data centers due to its reliability and relatively rapid deployment timelines of 18-24 months. Planned natural gas capacity increased from 11.1% in 2024 to 18.1% in 2026, and non-renewable additions surged by 71% from 2025-2026, while renewable growth flattened to just 2% over the same period. In the U.S., natural gas is the biggest source of electricity for data centers (40%), followed by renewables (24%), nuclear (20%), and coal (15%). However, renewables are projected to meet nearly half of the additional global demand, with natural gas and coal making up the rest. Small modular reactors (SMRs) are also gaining significant interest for their potential to provide carbon-free baseload power.

The Hidden Costs: Water, Land, and Carbon Footprint

Beyond the raw electricity demand, my research has illuminated other significant environmental impacts that the AI boom is creating. I found that data centers are voracious consumers of water, primarily for cooling server stacks. Even a mid-sized data center can consume as much water as a small town, while larger ones can require up to 5 million gallons daily, comparable to a city of 50,000 people. In the U.S. alone, AI-related data centers could require up to 32 billion gallons of water annually by 2028. This massive demand is a pressing problem, especially for communities already facing drought or depleting water supplies.

I've also observed the significant land footprint. Large data centers being built today can cover hundreds of acres, converting farmland or natural areas into impermeable surfaces and requiring new transmission line corridors. The environmental toll extends to carbon emissions as well. Despite efforts towards renewables, the current rate of AI growth could annually put 24 to 44 million metric tons of carbon dioxide into the atmosphere by 2030, which is the emissions equivalent of adding 5 to 10 million cars to U.S. roadways. This puts the AI industry's net-zero emissions targets out of reach without aggressive mitigation. While power consumption per AI task is declining due to efficiency improvements, the sheer increase in AI usage and energy-intensive applications like AI agents means overall consumption continues to climb.

A Geopolitical Energy Race

What I’ve come to understand is that the rise of AI is not just an energy story; it's a geopolitical one. I believe that nations increasingly view strategic control over compute, data, and digital infrastructure as central to national power. Semiconductor supply chains, cloud regions, and data corridors are becoming as politically significant as shipping routes or oil pipelines. Governments are moving rapidly to secure technological autonomy.

This creates what I call a "triple transition": the simultaneous transformation driven by advances in AI, the restructuring of global energy systems, and an accelerating geopolitical realignment. Countries that can provide secure, affordable, and rapid access to electricity will be best placed to benefit from AI development. I've seen that the U.S.'s sustained AI leadership may depend on its ability to address these energy demands effectively, as competitor countries like China and those in the Gulf are making significant state-led investments in AI infrastructure, sometimes detached from normal market logic, to accelerate their domestic AI efforts. For instance, Saudi Arabia's Vision 2030 strategy has already led to the creation of the Saudi Data and AI Authority. This global competition for AI compute resources and the energy to power them is reshaping strategic geographies and national priorities.

What This Means For Investors, Entrepreneurs, and Professionals

For investors, I see immense opportunities in the energy sector, particularly in infrastructure related to generation, transmission, and storage. Companies specializing in advanced cooling technologies, SMR development, and carbon capture solutions are poised for significant growth. I believe it's crucial to evaluate companies not just on their AI capabilities, but on the robustness and sustainability of their energy strategies. Caution is warranted regarding regulatory risks, as governments grapple with managing demand and ensuring grid stability, which could lead to policy shifts impacting profitability or creating stranded assets for utilities.

Entrepreneurs have a fertile ground for innovation. I envision a surge in demand for niche solutions in energy efficiency, advanced liquid cooling systems, and specialized microgrid development that can integrate various renewable and conventional sources. The paradox of "energy for AI and AI for energy" means that AI itself can be leveraged to optimize grid resilience and energy management, creating opportunities for AI-driven energy solutions. Furthermore, businesses focused on strengthening supply chains for critical electrical equipment, like transformers, will find a ready market.

For professionals, this era demands a new blend of skills. I believe there will be a growing need for electrical engineers, power systems experts, and data center architects who possess a deep understanding of energy infrastructure. Sustainability consultants, policy analysts, and project managers with expertise in large-scale energy and digital infrastructure development will be highly sought after. The complexity of these challenges means interdisciplinary collaboration will be key, requiring professionals who can bridge the gap between AI technology, energy systems, and environmental stewardship.

Bottom Line

I've concluded that the AI revolution is fundamentally an energy revolution, driving an unprecedented demand that traditional grids cannot meet alone. Big tech companies are stepping in to build their own power infrastructure, a move that is reshaping global energy markets, impacting environmental resources, and becoming a critical component of geopolitical power. I believe the future of AI hinges on innovative, integrated energy solutions that address not just power generation, but also grid resilience, resource efficiency, and sustainable development on a global scale.