Oracle Is Promising 2 Gigawatts of Renewables to Silence Stargate Data Centre Critics

Oracle has announced it will procure two gigawatts of renewable energy to offset the emissions from the Stargate AI data centre project it is co-developing with OpenAI and SoftBank. The move is a direct response to mounting pressure from local communities, environmental groups, and some government officials who have raised concerns about Stargate's energy footprint.

Stargate, announced earlier this year, is a 500 billion dollar initiative to build AI data centre infrastructure across the United States. It is one of the largest infrastructure bets in American tech history, and its energy requirements are correspondingly enormous. Data centres of the scale Stargate is planning consume power at levels that stress regional electricity grids and raise genuine questions about the sustainability of the AI build-out.

What Two Gigawatts Actually Means

Two gigawatts of renewable energy capacity is a significant number. To put it in context, the UK's Hornsea One offshore wind farm, which was the world's largest when it opened in 2019, has a capacity of 1.2 gigawatts. Oracle's commitment, if fulfilled, would represent more renewable energy capacity than that flagship project.

The critical caveat is that renewable energy commitments by large technology companies almost never work the way they sound. Most of these commitments are structured as power purchase agreements, where a company agrees to buy a volume of renewable electricity equivalent to its consumption, without any guarantee that the electricity actually powering the data centres at any given moment is coming from renewable sources. When data centres draw power at night or during low-wind periods, they're typically drawing from whatever the local grid provides, which includes fossil fuel generation.

Oracle has not specified the structure of its two gigawatt commitment, which leaves the actual environmental impact genuinely unclear. A 24/7 carbon-free energy commitment, where the company matches clean energy to its consumption hour by hour, is significantly more demanding and more impactful than a standard annual matching approach.

Why Stargate's Energy Use Is Politically Sensitive

The Stargate project has attracted opposition in several of the communities where data centres are planned. The objections range from concerns about water use for cooling, to the pressure large power consumers place on grid infrastructure that local residents and businesses depend on, to broader concerns about whether building enormous AI infrastructure is a good use of the country's clean energy capacity.

Some critics have noted the irony that data centres training AI models on climate data and environmental research are themselves major contributors to grid emissions. Others have focused on the more immediate and local impacts: data centres are large, loud, hot, and they can push up electricity prices in the areas they operate by increasing demand.

Oracle's renewables announcement is clearly designed to defuse at least some of this opposition. Whether it succeeds will depend on the specifics of how the commitment is structured and whether local communities find it credible.

The Broader Data Centre Energy Problem

Oracle is not alone in facing this scrutiny. Microsoft, Google, and Amazon have all made significant renewable energy commitments for their data centre operations, and all three have faced questions about whether their actual carbon impact matches their public commitments.

Microsoft in particular attracted attention in 2024 when its annual sustainability report showed its emissions had risen by 30% since 2020, largely driven by data centre construction and operation. The company attributed the increase to the steel and concrete used in building new facilities, as well as increased power consumption. It maintained its commitment to being carbon negative by 2030, but the data made that target look considerably more challenging.

The AI industry's energy problem is real and structural, not a communication problem. Training frontier models requires vast amounts of compute running for months. Inference, running the models for users around the clock, requires even more power in aggregate. The Stargate project is explicitly designed to build the infrastructure for this at national scale.

Two gigawatts of renewables is a real number. But it's also a number that will need to be tracked, verified, and held to account over time. Oracle announcing the commitment is the easy part. Actually delivering additionality, clean energy that would not have been built otherwise, is where these commitments either mean something or they don't.