In July 2026 Microsoft released its 2026 Environmental Sustainability Report covering fiscal year 2025. Total greenhouse gas emissions rose 25 percent year over year, from roughly 16 million to more than 20 million metric tons of carbon dioxide equivalent. The company attributed the rise primarily to rapid expansion of datacentre infrastructure for artificial intelligence and to a decision to pause the use of non-additional unbundled renewable energy certificates (energy attribute certificates, or EACs). It restated long-term goals to become carbon negative, water positive and zero waste by 2030, and framed the tension in its own words: sustainability solutions are not scaling fast enough to meet demand, and the tension is real and also productive.
Our view: Absolute climate targets break as a reporting calendar when CapEx-driven Scope 3 (capital goods) and EAC/additionality policy change in the same year. For HKEX issuers with AI/datacentre/tech growth exposure, the work is a measurement-control file for high-growth scenarios (boundary, CapEx GHG, EAC integrity, water/waste unevenness disclosed honestly) — not a narrative about one US tech peer.
Additionality, in energy-attribute accounting, means the clean-power claim is tied to projects that bring net-new carbon-free generation to the grid, rather than certificates that merely reallocate existing attributes without changing the physical supply. Scope 3 capital goods are the cradle-to-gate greenhouse gas emissions embedded in purchased capital equipment and construction — for AI buildouts, typically servers, cooling plant and datacentre fabric. A high-growth measurement-control file is a board-usable evidence pack that freezes boundary, CapEx GHG ownership, EAC integrity policy, and uneven environmental-pillar status while growth CapEx is still running, so Part D climate disclosure does not wait for annual-report prose to invent the numbers.
The problem for HKEX issuer teams
The failure mode we keep seeing is not “Microsoft is interesting.” It is that IR and sustainability teams treat a peer sustainability report as a narrative template: copy the ambition language, smooth the year when CapEx spikes, and hope the absolute target still looks coherent by drafting week. For issuers with AI, datacentre, cloud, hardware or other energy-and-materials intensive growth, that calendar is backwards.
Absolute targets assume a relatively stable activity base. CapEx-driven Scope 3 does not. When capital goods for new facilities land in the inventory in the same year the company upgrades EAC quality by dropping non-additional certificates, two legitimate control moves move the headline numbers violently in the same reporting period. Scope 3 still dominated Microsoft’s footprint at nearly 86 percent, with capital goods linked to datacentre construction accounting for most of the rise. Scope 2’s share of the total footprint jumped from under 2 percent to 13 percent after the pause of non-additional EACs. Matching 100 percent of annual electricity consumption with renewables and expanding contracted renewables to as much as 40 GW did not prevent the absolute total from rising. That is not a contradiction to “explain away.” It is a reporting-calendar collision.
Checkbox compliance looks like waiting for a neat story. Execution looks like freezing who owns CapEx GHG estimates, which EAC policy sits in the Scope 2 method note, and how water and waste results are allowed to diverge from the emissions headline without being buried.
How to read Microsoft’s disclosure choices as controls lessons
We read four choices as controls lessons, not as a scorecard for one company.
First, additionality over convenience. Discontinuing non-additional unbundled certificates and prioritising projects that bring net-new clean power raises near-term reported Scope 2 while improving the integrity of the claim. For HKEX Part D climate disclosure, the lesson is methodological honesty: if your market-based Scope 2 method changes, label the policy change in the same evidence file as the number. Do not let IR discover the method shift in drafting week.
Second, CapEx Scope 3 as the growth shock absorber. Capital goods will dominate for any issuer building physical infrastructure at scale. Deeper supplier data, design-for-lower-embodied-carbon, and named ownership of CapEx GHG estimates matter more than a smoother narrative. Inventory integrity belongs with ESG reporting services discipline; who signs CapEx estimates and method changes belongs with internal control advisory logic.
Third, uneven pillars disclosed honestly. Microsoft reported that in FY2025 it replenished more water globally than it withdrew for the first time — more than 14 million cubic meters — on the path to a 2030 water-positive target. On waste, it reported reusing or recycling 92 percent of decommissioned servers and components for a second consecutive year, diverting 90.5 percent of construction and demolition waste from landfill, expanding Circular Centers to seven facilities, and cutting single-use plastics to 0.07 percent of primary product packaging by weight. Emissions moved the wrong way; water and waste did not. Smoothing that into one uniform “progress” paragraph is how credibility breaks. Part D evidence should allow pillar-level divergence to stay visible.
Fourth, counterfactuals labelled outside the inventory. Microsoft published a modelled comparison of emissions without four interventions (carbon-free electricity purchases, sustainable aviation and marine fuel, Xbox efficiency, Surface device decarbonisation) and was explicit that the comparison is directional and sits outside the official greenhouse gas inventory — not an additive avoided-emissions claim. That labelling is a governance control. For issuers writing HKEX climate disclosure, keep directional “what if” stories out of the inventory totals, and keep bilingual listing-rule terms consistent via financial translation as a control, not cosmetics. Where transition CapEx assumptions touch recoverable amounts or fair-value narratives, keep business and property valuation assumptions on the same evidence spine.
Carbon dioxide removal remains part of Microsoft’s long-term framing through multi-year offtake agreements across engineered and nature-based pathways. Removal can address residuals. It cannot substitute for owned measurement of CapEx and operational emissions while the buildout is running.
Implications for HKEX Part D climate disclosure evidence
HKEX Part D climate disclosure asks issuers for decision-useful climate information — including greenhouse gas measurement, risks and opportunities with financial-effect links, and scenario and transition thinking, subject to the Code’s own phase-in and relief architecture. The Microsoft episode is not a template for copying US tech prose into a Hong Kong ESG report. It is a stress test of whether your evidence file survives high-growth CapEx.
Three implications follow for issuer teams.
If your strategy includes AI, datacentre, semiconductor, cloud or other CapEx-heavy growth, absolute-target pathways must be stress-tested against a high-growth CapEx case before the year closes. A pathway that only works in a flat CapEx year is not a pathway; it is a drafting hope.
If you use market-based Scope 2 instruments, document additionality policy as a control attribute: what quality of EACs or power purchase agreements is allowed, what changed this year, and how that change moves the reported total. Quiet upgrades in integrity that explode Scope 2 without a method note read as surprises, not as governance.
If water, waste or other environmental pillars outperform while absolute emissions rise, disclose the unevenness. Markets and relationship banks can read a coherent evidence file with divergent pillars. They struggle with a single smoothed narrative that contradicts the CapEx calendar they already see in the financial statements.
What to do now
Build one high-growth measurement-control file while CapEx is still live. Freeze organisational boundary and consolidation approach for climate figures. Map Scope 3 capital-goods ownership (who estimates, who signs, which supplier data windows are open). Record EAC / market-based Scope 2 policy and any additionality upgrade or pause in the same file as the Scope 2 total. Separate inventory totals from any directional counterfactual, with an explicit “outside inventory” label. Allow water and waste (or other material pillars) to report their own trajectory beside emissions, without forced uniformity. Name owners for scenario and transition assumptions that embed AI or datacentre growth.
Stop treating a peer’s sustainability report as the work. The work is whether your Part D evidence can survive the year CapEx and instrument policy both move.
FAQs
- Why did Microsoft’s FY2025 greenhouse gas emissions rise 25 percent?Per Microsoft’s 2026 Environmental Sustainability Report (FY2025), total emissions rose about 25 percent year over year, from roughly 16 million to more than 20 million metric tons of CO₂e, driven primarily by AI datacentre CapEx (Scope 3 capital goods) and by pausing non-additional unbundled energy attribute certificates, which increased the reported Scope 2 share of the footprint.
- What does EAC additionality mean for climate disclosure?Additionality means clean-power claims are tied to net-new carbon-free generation rather than certificates that only reallocate existing attributes. Raising additionality standards can increase near-term reported market-based Scope 2 even when renewable matching remains high; the control task is to label the policy change beside the number.
- What should HKEX issuers take from this for Part D climate disclosure?Treat the episode as a controls lesson for high-growth CapEx: freeze boundary, CapEx GHG ownership, EAC integrity policy, and uneven pillar evidence in a measurement-control file while the growth year is running. Do not wait for annual-report drafting to reconstruct method changes or capital-goods estimates.
- Should modelled “avoided emissions” sit inside the GHG inventory?Microsoft labelled its counterfactual comparison as directional and outside the official inventory rather than as an additive avoided-emissions claim. For HKEX climate disclosure, keep inventory totals and directional what-if analyses in separate, clearly labelled layers so certainty is not overstated.
- What is a high-growth measurement-control file?A short board-usable evidence pack that freezes climate boundary, CapEx Scope 3 ownership, EAC/additionality policy, inventory-versus-counterfactual labelling, uneven environmental-pillar status, and named scenario owners while growth CapEx is still underway. It is a measurement-control artefact for Part D evidence, not a peer-narrative template.
Conclusion
Microsoft’s 25 percent emissions rise is useful to HKEX issuer teams only if it is read as a reporting-calendar break: CapEx-driven Scope 3 capital goods and EAC additionality policy moved in the same year, while water and waste progressed on different tracks, and a directional counterfactual was kept outside the inventory. Absolute climate targets do not fail because ambition was wrongly worded. They fail as a calendar when growth CapEx and instrument integrity are not owned as controls. For issuers with AI or datacentre exposure, the executable next step is a high-growth measurement-control file for HKEX Part D climate disclosure — not another narrative about one US tech peer.
Edited / written by: Amanda Cheung, Founder, AMMO Group
