The Energy Crisis Behind the AI Boom in the United States 

By Isabella Arrigo

The infrastructure underpinning artificial intelligence doesn’t live in the cloud – it lives in buildings that consume enormous amounts of electricity. In the United States, hyperscalers need more and more power to build their data centers. The main archetype today remains the hyperscale, grid-connected data center – massive facilities drawing power from public utilities, accounting for the majority of the market. But a more revolutionary model is emerging: off-grid data centers powered by onsite generation, built to bypass a grid that cannot keep up.

A Grid Under Pressure

The numbers are striking. US data center power demand is projected to climb from 31 GW in 2025 to 66 GW by 2027, according to Goldman Sachs Research, more than doubling in two years.

The problem is simple. The demand for computing is doubling roughly every few months, driven by AI training and inference workloads. Efficiency improvements, in hardware, cooling, or power design, tend to double on a two to three year cycle. The gap between demand velocity and efficiency velocity is what’s forcing a rethink of how data centers are powered.

As a result, many new data centers are choosing to generate their own power. According to Bloom Energy’s 2025 Data Center Power Report, 27% of data center facilities are expected to be powered by onsite generation by 2030, up from 13% a year prior. Most of this onsite capacity is natural gas. Behind-the-meter gas plants can be deployed in as little as 18 months – a competitive advantage when utility grid connections in key markets are running one to two years behind developer expectations. It solves the immediate problem. Whether it’s the right solution is a different question.

Who Is Doing It Cleanly?

Among the major hyperscalers – Microsoft, Google, Meta, and Amazon – approaches to sustainability vary considerably, and Google stands out as appearing to care the most.

Rather than buying renewable energy certificates to offset consumption, Google matches its electricity use to carbon-free energy on an hourly basis, a more rigorous standard. More strikingly, Google has crossed 1 GW of demand response capacity across its US data centers, meaning it can shift workloads to ease grid pressure during peak hours. This also positions Google as a grid services provider. One that unlocks capacity for everyone.

What This Means for Emerging Markets Like Turkey

The dynamics playing out in the US are arriving in markets around the world. Turkey is an example. The country’s data center market is projected to grow from $715 million in 2025 to $1.79 billion by 2031, driven by AI adoption, cloud migration and Turkey’s strategic position between Europe and the Middle East. Global operators have already taken notice. DAMAC Digital and Vodafone committed $100 million to a new Izmir campus in 2024, and Khazna Data Centers announced a 100 MW facility near Ankara in 2025. Yet the same challenge applies: Turkey’s renewable generation sites are in remote areas, while demand is concentrated in the urban northwest and a $10 billion grid upgrade program won’t be completed until the early 2030s.

Conclusion

The AI boom is, above all, an energy story. The innovation happening inside data centers – smarter chips, cleaner procurement and smarter energy use – is impressive. But the gap between the pace of demand and the pace of grid development is real, and no single solution is enough.

About Isabella Arrigo

Isabella Arrigo

Isabella Arrigo holds a B.A. in International Relations and an M.A. in Earth Systems & Sustainability from Stanford University. She currently works as a Business Operations Associate at Verne, an energy startup based in San Francisco providing behind-the-meter power generation for data centers. Before Verne, she worked on the investing and responsible investing teams at leading private equity firms (Arsenal Capital Partners, Makena Capital Management, Keensight Capital) and on climate mitigation and adaption projects at the World Bank within the Development Impact Group. Through her work, she is committed to accelerating the deployment of more sustainable energy solutions.