Microsoft considers shelving 2030 clean energy target as AI power draws climb

TL;DR:

• Microsoft is considering delaying or abandoning its 2030 goal of matching its entire hourly electricity use with renewable purchases, Bloomberg News reported on 6 May, citing people familiar with the matter; no final decision has been made.
• The goal was set before the AI era; Microsoft is now spending hundreds of billions of dollars on AI infrastructure, with multiple new data centres expected to have multi-gigawatt capacities — a single gigawatt powers roughly 750,000 US homes.
• Resultsense view: this is the second corporate climbdown signal of the week on AI’s hidden costs, after the Financial Stability Board flagged AI-related private credit on Tuesday. UK readers should expect Whitehall and the City to start asking the same questions of UK-located hyperscaler capacity within the year.

A Microsoft spokesperson told Reuters the company “continues to look for opportunities” to maintain the matching goal, citing recent agreements with We Energies to bring 1.2 gigawatts of carbon-free energy projects in Wisconsin onto the grid, including solar and battery installations expected to begin coming online in December 2028. The implication is that the targets remain ambitious; the operative question is whether AI power demand has outrun even those commitments.

Why the goal is under strain

Microsoft, like Amazon and Alphabet, is spending hundreds of billions of dollars to build AI infrastructure for products like its Copilot assistant and Azure cloud service. The new data-centre footprint expected over the next several years includes multi-gigawatt facilities — a step change from prior generations. To deliver that capacity at the speed AI buyers want, hyperscalers have signed nuclear, natural gas and renewable deals; Microsoft itself agreed in 2024 with Constellation Energy to help resurrect a unit of the Three Mile Island nuclear plant in Pennsylvania.

Several industry executives have publicly noted that natural gas is faster and easier to deploy than renewables at the gigawatt scale required, putting pressure on the climate maths underpinning hourly-matched targets specifically — which require renewable supply at every hour, not just on a net annual basis.

Cross-source context

This sits inside a wider AI-and-energy story building over the past month. The Financial Stability Board on 6 May warned that the private credit industry financing the AI build-out poses systemic financial risk. UK chip designer Arm forecast Q1 revenue above expectations the same day, citing AI data-centre demand. Anthropic, also this week, secured the entire 300-megawatt Colossus 1 facility from SpaceX, and pledged to cover any consumer electricity price increases caused by its US data centres.

UK and grid context

Britain’s data-centre footprint is materially smaller than the US hyperscaler clusters in question, but the pattern is now visible domestically: large new AI-capable facilities are being proposed across the South East and Midlands, and grid-connection queues have become a UK policy issue in their own right. The Microsoft signal will sharpen UK political and regulatory questions about the carbon, water and grid implications of any UK AI infrastructure plan — questions Liz Kendall’s June AI hardware plan will need to address.

Looking forward

A formal Microsoft decision has not been made and may not be made publicly. The more useful indicator over the next 12 months is whether other hyperscalers explicitly soften their hourly-matched commitments. The credibility of the corporate net-zero AI narrative depends on it.