Microsoft has reported a 25% annual increase in its combined Scope 1, 2, and 3 greenhouse gas emissions as rapid expansion of artificial intelligence infrastructure drives higher demand for data centres, electricity, construction materials, and computing equipment.
The technology group’s emissions reached approximately 20 million tonnes of carbon dioxide equivalent during its latest reporting period. Its total footprint now stands substantially above the 2020 baseline against which Microsoft measures progress towards becoming carbon negative by 2030.
Construction and operation of data centres accounted for much of the increase. Microsoft is expanding its global estate to support cloud services and energy intensive artificial intelligence workloads, requiring new buildings, servers, cooling systems, network equipment, and power capacity.
A large share of the resulting carbon sits within the supply chain rather than Microsoft’s direct operations. Semiconductor manufacturing, steel, concrete, transport, construction, and replacement of computing equipment contribute emissions before a facility begins serving customers.
The reported increase also reflects a change in renewable energy accounting. Microsoft has begun reducing its use of some unbundled renewable energy certificates that do not directly support the construction of new clean generation within the grids serving its facilities.
Annual electricity consumption can be matched with certificates while data centres continue drawing power from a grid that relies partly on fossil fuels. Contracts linked to additional wind, solar, storage, nuclear, or other low carbon capacity may exert a greater effect on the physical system, although those projects take time to finance, permit, connect, and build.
Microsoft continues to pursue environmental commitments covering carbon, water, waste, and land. Progress in recycling and conservation sits alongside rising emissions, illustrating the difficulty of reducing absolute impact while the underlying business and infrastructure footprint grows quickly.
Artificial intelligence has increased the scale of that challenge. Training and operating large models requires dense clusters of specialist chips, substantial cooling, and continuous electricity. Demand continues after a model has been trained because widespread use across search, workplace software, coding, customer service, and analytics creates recurring workloads.
Efficiency improvements do not necessarily reduce total consumption. A model may require less energy for each task, while overall electricity use still rises because more customers, products, and employees use it more frequently.
Scope 3 emissions are particularly difficult to control because they arise from suppliers, contractors, logistics, capital goods, and customer activity. Microsoft can set procurement standards and sign long-term agreements, but it does not directly manage every factory, construction site, or transport route within its value chain.
Disclosure is receiving closer examination as companies promote ambitious climate goals alongside substantial growth. An Advertising Standards Authority ruling has already raised the evidential standard for environmental claims, increasing pressure on companies to explain accounting boundaries, certificates, offsets, and the effect of expansion.
Recognising higher emissions after reducing reliance on weaker certificates may improve the credibility of Microsoft’s reporting, although it also exposes the distance between its current trajectory and an absolute reduction.
The company must add clean generation, improve energy efficiency, reduce the carbon intensity of construction, and influence suppliers faster than demand for computing grows. Each of those changes depends on industries with their own investment cycles and infrastructure constraints.
Other technology groups face similar pressures. Cloud and artificial intelligence providers are competing for grid connections, long-term power contracts, specialist equipment, water access, and suitable land, while some proposed data centres are being delayed because local electricity networks cannot accommodate them.
Corporate customers will encounter a portion of these emissions through their own Scope 3 reporting. Transferring workloads from internal servers to the cloud can reduce the need for on-site infrastructure, but it does not remove the environmental impact of computing.
Procurement teams may consequently seek more detailed information about the energy and carbon intensity of cloud services. Location, workload scheduling, hardware utilisation, and the electricity mix serving a data centre can all affect the footprint associated with the same digital task.
Investors and regulators are also likely to distinguish between efficiency per unit of computing and the company’s total emissions. Relative improvements can coexist with a rising absolute footprint, particularly during a period of rapid infrastructure construction.
The financial consequences extend beyond environmental reporting. Clean power contracts, grid upgrades, efficient cooling, lower carbon materials, and supplier transition programmes require capital, while delayed access to electricity can postpone revenue from new facilities.
Microsoft’s 25% increase reflects a broader tension across the technology sector, where artificial intelligence investment is advancing more quickly than many energy systems and industrial supply chains can decarbonise.
Subsequent reports will show whether the current rise represents a peak produced by intensive construction or the beginning of a longer trend. Absolute emissions, the quality of new power procurement, and progress among suppliers will provide a clearer measure of performance than efficiency gains alone.
You must be logged in to post a comment.