TL;DR The same AI systems companies are using to advance their sustainability goals are generating emissions that most of them have not yet learned to disclose, measure, or explain. Microsoft’s carbon footprint jumped 25% in 2025, driven primarily by AI data centre expansion. Google and Amazon reported similar double-digit increases. If your clients are using AI at scale and claiming sustainability credentials, their messaging has a structural vulnerability that will surface under scrutiny. PR firms that understand this now are ahead of a very consequential story.
There is a contradiction sitting in the middle of most corporate sustainability communications right now, and it is going to attract attention. Companies across sectors are using AI to monitor energy consumption, optimize supply chains, automate emissions reporting, and build the data infrastructure that makes credible ESG disclosure possible. According to Deloitte’s 2025 Global C-Suite Sustainability Report, 81% of executives already report using AI to advance their sustainability objectives. The efficiency gains are real. The case for AI as a sustainability tool is credible.
The same companies are not yet accounting for what those AI systems themselves emit. Microsoft’s 2026 Environmental Sustainability Report disclosed a 25% jump in total carbon emissions year-over-year, driven primarily by AI data centre expansion. Google reported an 18% increase, Amazon a 16% rise. These are the companies building AI infrastructure. But AI’s carbon cost is not confined to infrastructure providers. Every company paying for AI API access, enterprise seat licences, or AI-embedded SaaS tools is purchasing emissions they are not yet required to attribute, and in many jurisdictions will soon be required to disclose.
For PR firms, this is not a technical problem to hand off to the sustainability team. It is a communications strategy problem. And it needs to be addressed before a journalist or regulator raises it.
Why AI’s Emissions Are Different from Other Digital Carbon
PR practitioners are generally comfortable advising clients on digital carbon: the emissions associated with website hosting, cloud storage, employee devices, video conferencing. These are well-understood, routinely included in Scope 2 calculations, and increasingly addressed through renewable energy procurement or carbon offsets. AI workloads are categorically different in scale. According to the International AI Safety Report, AI consumes between 10% and 28% of global data centre energy capacity, and AI energy demand is projected to double or more by 2026. Data centres and data transmission already account for an estimated 1% of global energy-related greenhouse gas emissions, with AI responsible for a disproportionate and growing share of that figure.
The carbon footprint of AI systems globally was estimated at between 32.6 and 79.7 million tons of CO2 equivalent in 2025, with water consumption potentially reaching 312 to 764 billion litres. In Ireland, data centres already consume 21% of national electricity, with the IEA estimating that share could rise to 32% by 2026. These are not marginal numbers.
What makes AI emissions particularly difficult for corporate sustainability communicators is the attribution problem. When a company runs an AI prompt through a third-party API, the emissions from that computation are Scope 3 Category 1 under the GHG Protocol, classified as purchased goods and services. Most corporate Scope 3 inventories do not yet include AI API costs as a tracked line item. Under the EU’s CSRD, which requires disclosure of all material Scope 3 categories, this gap is not a technicality. It is a compliance issue that will surface in audits.
The Messaging Vulnerability This Creates
Consider the situation from the outside looking in. A company publishes a sustainability report claiming progress on Scope 1 and 2 emissions, reduced travel, renewable energy procurement, a supplier code of conduct. The report also highlights the company’s investment in AI-powered monitoring tools to track supply chain emissions in real time. The implied message is that AI is part of the sustainability solution. A journalist or activist with access to the company’s technology stack and a basic understanding of AI carbon accounting can ask: have you disclosed the emissions from the AI tools you are using? If the answer is no, the story writes itself. ‘Company touts AI as sustainability tool while failing to account for AI’s own carbon footprint’ is not a hypothetical. It is the kind of investigation environmental reporters are actively developing as CSRD disclosure requirements bring Scope 3 data into sharper relief.
The vulnerability is compounded by the opacity of AI emissions data. A 2025 Watershed survey of corporate sustainability teams found that most teams spend 40 to 50% of their time on data collection, cleaning, and reporting rather than on decarbonisation strategy. AI API costs, which have moved from negligible to material within three years, are rarely in the data collection scope. The gap is not malicious. It is a planning failure, and it is one that PR firms are better positioned to identify than sustainability teams who are focused on the operational reporting process.
How Microsoft, Google, and Amazon Changed the Reference Point
The big tech disclosures of mid-2026 are important for PR practitioners not because of their scale but because of how they changed the reference point for corporate ESG communications. Before July 2026, the implicit assumption in most corporate sustainability communications was that AI and sustainability were aligned interests: AI would help companies measure, reduce, and report emissions more accurately, and the efficiency gains from AI would net out the energy costs. The Microsoft and Google reports collapsed that assumption publicly. A company that has built its sustainability narrative around AI-enabled efficiency now has to account for the fact that the leading AI providers are themselves reporting sharp emissions increases driven by the same infrastructure those efficiency tools run on.
Microsoft’s response is worth studying as a communications model. Rather than minimising the 25% emissions increase, the company’s sustainability report contextualised it: the jump was driven by a deliberate decision to stop using unbundled renewable energy certificates and to invest in genuinely additional carbon-free energy infrastructure instead. The short-term increase was framed as a consequence of a principled long-term decision. That framing was not spin. It was honest, specific, and substantiated by the company’s carbon removal portfolio, where Microsoft accounts for roughly 78.5% of all disclosed durable CDR volumes contracted globally. The disclosure held up under scrutiny because the numbers behind it were real.
For corporate communications teams whose clients are not Microsoft, the lesson is not to replicate the CDR strategy. It is to replicate the disclosure discipline: know what the numbers are, own the narrative before someone else does, and provide specific context rather than vague reassurance.
What PR Firms Should Be Doing Right Now
Audit the AI carbon gap before anyone else does
The first step is the most uncomfortable one. PR firms advising clients with sustainability credentials need to identify whether those clients have AI tools embedded in their operations, whether the emissions from those tools have been estimated, and whether they are included in any existing Scope 3 disclosure. For most clients, the answer to the last two questions will be no. That is a vulnerability, not a crime, and it is far better for a PR firm to surface it internally than for it to surface externally. The audit does not need to be comprehensive. It needs to be honest enough to identify the gap before it becomes a headline.Help clients get ahead of the disclosure curve
Under the EU CSRD framework, material Scope 3 categories must be disclosed. AI API costs are a purchased good or service under Category 1. As AI adoption scales across organisations, the materiality threshold for AI-related emissions will be crossed by an increasing number of companies. The PR firm’s job is not to produce the accounting methodology. It is to make sure the client’s sustainability communications team is working with finance and technology on this before the next reporting cycle, and to advise on how the first disclosure of AI-related emissions should be positioned when it comes.Reframe AI’s role in the sustainability narrative
The narrative that AI is purely a sustainability enabler is now incomplete. A more sophisticated and more defensible narrative acknowledges the dual role: AI as a tool for sustainability optimisation, and AI as a source of emissions that needs to be managed and disclosed. Companies that proactively adopt this framing are ahead of the story rather than behind it. For clients in technology, financial services, and manufacturing sectors where AI adoption is heavy, this reframing is not a defensive move. It is a credibility builder with sophisticated investors and ESG-literate journalists who already know the contradiction exists.Distinguish between AI use emissions and AI infrastructure emissions
One communications nuance that matters significantly: there is a difference between the emissions generated by a company’s own AI infrastructure (Scope 1 and 2) and the emissions generated by third-party AI services the company purchases (Scope 3 Category 1). Most companies outside the hyperscaler tier are in the second category. The communications challenge is different. The argument is not ‘our AI systems are sustainable’ but ‘we are building a methodology to account for AI usage emissions in our Scope 3 reporting and will disclose that estimate in the next annual report.’ That is a forward-looking, process-oriented disclosure that is honest about the current gap without pretending the gap does not exist.AI Carbon Disclosure: Where Different Company Types Stand in 2026
| Company Type | Likely AI Emissions Exposure | Current Disclosure Gap | Recommended Communications Posture |
|---|---|---|---|
| AI infrastructure providers (hyperscalers) | High: Scope 1, 2, and supply chain Scope 3 | Partially addressed; still significant opacity on AI vs. non-AI workload attribution | Proactive annual disclosure with workload breakdown; Microsoft model is current benchmark |
| Enterprise AI users (finance, manufacturing, logistics) | Medium-high: material Scope 3 Category 1 exposure from API and SaaS costs | Largely unaddressed; most Scope 3 inventories predate material AI spend | Commission AI emissions estimate; build into next reporting cycle; disclose methodology proactively |
| SMEs using AI-enabled SaaS tools | Low-medium: individually small but cumulatively material at sector level | Typically unaware that AI embedded in standard software tools carries attribution | Include AI-enabled software in technology carbon audit; incorporate into SME sustainability reporting as sector standards develop |
| Companies using AI for ESG data management | Medium: the tools used to measure sustainability have their own footprint | Ironic but common gap: AI carbon not disclosed in the same reports AI helps produce | Acknowledge AI tools in ESG methodology disclosure; estimate and include usage emissions |
Frequently Asked Questions
Do companies need to disclose AI-related emissions in their ESG reports?
Under the EU CSRD, companies must disclose all material Scope 3 categories. AI API access, enterprise AI licences, and AI-embedded SaaS subscriptions are Scope 3 Category 1 (purchased goods and services). As AI spend becomes material, the disclosure obligation follows. In the US, the SEC’s climate disclosure rules similarly require material emissions to be reported. Companies with significant AI tool usage should not assume their current Scope 3 inventory adequately captures AI-related emissions; most do not yet include these costs as a tracked category.Why did Microsoft, Google, and Amazon all report emissions increases in 2025-2026?
All three companies reported that AI data centre expansion was the primary driver of increased emissions. Microsoft’s 25% year-over-year increase was explicitly attributed to infrastructure buildout and a decision to stop using non-additional renewable energy certificates. Google’s 18% increase reflected supply chain activities supporting rapid business expansion, and Amazon’s 16% rise similarly reflected infrastructure investment. These disclosures represent the most significant public acknowledgement to date that AI infrastructure growth is creating material tension with corporate sustainability commitments.What is Scope 3 Category 1 and why does it matter for AI emissions?
Scope 3 Category 1 under the GHG Protocol covers purchased goods and services, meaning the emissions generated in the production of what a company buys. Every AI API call, enterprise AI subscription, and AI-embedded software licence generates electricity consumption, and therefore emissions, at the data centre level. Those emissions are attributable to the companies buying the services under Category 1. As AI spend grows from negligible to material within corporate budgets, the Scope 3 Category 1 emissions from AI usage become a disclosure obligation under CSRD and increasingly under voluntary frameworks like the SBTi.How should a PR agency advise a client whose AI emissions they cannot yet quantify?
Honestly, and with a forward-looking process commitment. The right communications posture is to acknowledge that AI usage emissions are not yet fully captured in the current Scope 3 inventory, commit to developing a methodology for the next reporting cycle, and provide whatever estimate is possible in the interim. This approach is substantively better than silence, which implies either ignorance of the issue or deliberate omission. Methodological transparency, even when the data is incomplete, builds credibility with the ESG analysts and investors who are the primary audience for this disclosure.Is there a risk of greenwashing in claiming AI as a sustainability tool?
Yes, and it is a growing one. Companies that position AI as central to their sustainability strategy without disclosing AI’s own emissions create a potential greenwashing exposure, because the implied claim, that their sustainability approach is more effective because of AI, is incomplete without the counter-disclosure. This does not mean companies should stop using AI for sustainability purposes. The efficiency gains are real and documented. It means the communications around AI’s sustainability role need to include an honest account of AI’s environmental cost alongside the efficiency benefits it delivers.The AI carbon story is not going away. It is going to get louder as CSRD enforcement deepens, as more tech companies publish sustainability reports that show the same emissions trends Microsoft, Google, and Amazon have already disclosed, and as journalists and activists develop the literacy to identify which corporate sustainability claims are built on incomplete Scope 3 accounting. PR firms that understand this dynamic now, and help clients get ahead of it, are doing genuine strategic work. Those that wait for the headline to appear will be managing a harder problem with less time and less credibility.
Madchatter Brand Solutions is a Worldcom PR Group partner based in Mumbai, India, specialising in strategic communications for deep tech, B2B, fintech, and technology-adjacent businesses.
