Powering data centres: AI leadership depends on control of power supply

Google's Intersect Power deal shows how AI's explosive energy needs are pushing Big Tech from buying clean power to owning it-reshaping grids, climate goals and regulation

Google’s reportedly planned $4.75 billion acquisition of Intersect Power will not just be another corporate clean-energy deal. Data centres powering large-language models now consume city-sized amounts of energy, and the constraint is no longer only chips or capital but also access to reliable power at scale. By buying a major renewables developer, Google is moving beyond being a buyer to becoming an energy actor in its own right. For years, technology companies relied mainly on power-purchase agreements (PPAs). In this case, ownership will give Google the ability to build generation capacity to power new data-centre campuses. Intersect has an array of solar and battery-storage projects concentrated in Texas, a fast-emerging data-centre hotspot. 

The move also reflects a deeper tension within Big Tech. Expansion in artificial intelligence (AI) is colliding with climate commitments. In its latest annual environmental report, Google has admitted that the company’s overall emission was up 51 per cent relative to a 2019 baseline. Securing clean energy is, therefore, not only about reliability of power but also a pathway towards net-zero pledges. Intersect fits into a broader strategy where Google is building a diversified portfolio of energy solutions, including renewables, storage, geothermal, nuclear, and demand-response programmes that allow data centres to reduce load during periods of grid stress. The company has also been aggressively harnessing offshore wind power through some of the world’s largest corporate PPAs. Deals include 478 megawatts (Mw) in the Netherlands, 495 Mw in Taiwan, and additional capacities in Germany and Scotland. These agreements are central to Google’s goal of operating on carbon-free energy across every grid it uses by 2030. 

Google’s Intersect deal must be seen in the context of recent developments in the United States, where the Donald Trump administration and several governors of the country’s Northeastern states are pushing for an emergency 15-year wholesale electricity auction that would compel technology companies to fund new power plants. The idea is simple. Households and small businesses should not bear higher electricity bills because of the burgeoning energy requirements of hyperscale data centres. 

Yet building round-the-clock clean power remains expensive. Solar and wind require storage or backup generation. If large firms increasingly self-supply and bypass public grids, there is also a risk that the costs of maintaining shared networks get shifted to other consumers. Regulators will have to balance investment incentives with fairness. For India, the implications are immediate. Data-centre hubs are expanding rapidly across Maharashtra, Tamil Nadu, Gujarat, and Telangana. These facilities will not be marginal additions to demand; they are large, concentrated loads that can strain local distribution networks and require major transmission planning. India’s power sector has made progress in electrification and renewable deployment, but discom finances remain fragile and cross-subsidy structures depend heavily on large industrial consumers. India also lacks mature demand-response markets, which could allow a flexible curtailment of data-centre loads during system stress. California’s demand-response auction mechanism (DRAM) provides a useful benchmark in this context. As AI infrastructure grows, India will need clearer rules on who pays for grid upgrades, how storage is scaled up, and how renewable integration keeps pace with industrial demand.