2024 progress on the climate pledges of tech companies. And then there came AI.
The emissions overview in 2024 (with data from 2021–2023) and illustrating the respective trend in terms of emissions reduction 🟢, stagnation 🟡, or growth 🛑 presents a concerning trajectory. Last year, only two major tech companies were still increasing their absolute GHG emissions and the rest seemed to slowly stir in a positive direction — even if still falling shorting of their climate commitments and required reductions. Now, the “AI-fication” of everything has uprooted (and shattered) much of the desperately needed reductions. Climate commitments? Not on the current trajectory.
Emissions data 2021, 2022, 2023. Amount in million metric tons of carbon dioxide equivalent (mtCO2e) and the respective trajectory:
Alibaba
Alibaba’s 2024 ESG report puts emissions at 12.32 million mtCO2e across all three scopes (in the appendix). As before, the report is 200 pages long and, in its tone, clearly geared towards marketing rather than increasing transparency around the measurable impact of their actions. Data is challenging to extract and in the narrative part only accounts for 4.449 million tons of CO2e (MmtCO2e) — net and for scope 1 and 2 only. In terms of operations, this is put at a 5% reduction compared to 2022. Like last year, scope 3 emissions are presented as an intensity factor per million spent only, not in absolute numbers. In 2023, Alibaba accounts for 8.1 tons per million renminbi (RMB, the official currency in China) in revenue.
While the metrics table is more granular than previous years, it is unclear which data is measured, which estimated, and which accounted for only after offsets. Hence, the reported data should be taken with a grain of salt.
Amazon
Amazon’s latest GHG emissions report is 98 pages long and claims to account for a slight reduction in overall emissions, 3% in absolute numbers to now 68.82 million mtCO2e for 2023 (but still 13% above 2020 emissions and not on par to reach net-zero emissions by 2040). Critically, scope 1 emissions, those the company directly controls, have seen a 7% increase compared to last year and now account for app. 21% of their overall carbon footprint.
Note also that AWS services are not covered by this, which make up around 17% of their revenue, there is separate summary report available for AWS, only that emissions data for their cloud services cannot be extracted from it. To be clear: Purchasing renewable energy certificates to cover data center electricity does not reflect their full value chain emissions.
Pointing to business growth and the “nature of their business”, Amazon also reports that to date, they focused on expanding renewable energy; going forward, they need to “leverage additional carbon-free energy options — such as nuclear — to support our continued growth and enable us to develop and deploy new technologies such as AI” (p. 12, emphasis mine).
In other words, not only do we need a lot more transparency around the impact of Amazon’s cloud infrastructure and AWS services, but we also need a detailed breakdown of their AI investments. As both will determine whether Amazon’s climate pledge is hard, or impossible, to fulfil.
Apple
Apple’s latest environmental report is 113 pages long and accounts for 16.1 million mtCO2e in gross emissions for 2023, which amounts to a 22% reduction compared to 2022.
59% of those come from product manufacturing, 29% from product use and 9% from product transport. A model student in terms of transparency so far and the only company that continues to meet their reduction goals: This trend is slowly coming under pressure, and they have entered their offsetting phase. In its report, Apple points to the limits of their efforts: “not all emissions can be avoided or reduced with existing solutions. And some existing solutions will require greater effort from industry and government to scale them before broader commercial adoption becomes feasible.” (p. 10) In other words: Unless the system changes, we’re at our peak effort.
Google (Alphabet)
On 86 pages, Google reports 14.3 million mtCO2e in gross emissions for 2023 for Google only, not Alphabet and all its subsidiaries, including YouTube, Jigsaw, and more. This presents a 13% increase in absolute emissions.
The main driver for their increased emissions: electricity needs required by AI. Just to illustrate this: In 2023, Google consumed 25.3 TWh in electricity, 3.6 TWh more than the year before — an excess that could power as much as 1.2 million homes for a year.
Interestingly, their entire report is drafted to underscore the importance of AI for Google as a company and its growth opportunities, while framing AI as not just one, but the crucial tool to reduce global emissions. This can’t make up for the irony, though, that the pathway to reducing AI-related impacts remains unclear (as Google admits on p. 12). As is, these emissions are unlikely to be scaled back in the coming years, and without further scrutiny we will continue to see emissions soar.
Meta
Meta has not released their 2024 report yet. The latest data index available is from July 2023.
Microsoft
Microsoft’s 2024 environmental report accounts for 17.16 million mtCO2e, which need to be extracted from the 23-page appendix following an 88-page long report. Around 96.5% of their emissions fall under scope 3 with capital goods (38%), purchased goods and services (36%), and use of products (14%) topping the list. Note also that their overall emissions follow the market-based accounting approach, though the location-based approach would add another 7 million mtCO2e according to their data.
Their electricity usage has gone up by 5.4 TWh to 23.6 TWh in 2023 — and, to no surprise, this is first and foremost a result of Microsoft’s investment in AI applications, their energy, and data center needs. This growth is also reflected in increased water usage. As Microsoft will double down on its AI trajectory, their carbon negative pledge will move further and further out of reach.
Tencent
In 2023, Tencent accounts for 5.8 million mtCO2e, though the data is not (yet) available in a comparable table with metrics or indicators for their year over year change. The number is shared in a high-level visual in their overall ESG report (p. 14). In general, this indicates a slight upward, rather than a downward trend.
What this tell us?
The push towards more corporate responsibility is at a juncture. In recent years, businesses realised that markets would falter and break-away as more and more regions and people are feeling the consequences of the poly-crises (climate emergency, wars, economic pressures, pandemics, and more). Hence, companies pledged to assume responsibility and do their part in mitigating the worst impacts — in their own businesses’ interests.
Now, existing climate commitments are coming under pressure — and not because of failures to design a reasonable reduction pathway, but because pressures to increase revenue are (yet again) taking precedent. This time because the magic ingredient is Artificial Intelligence. Whether there is an actual problem that can be solved with AI applications or not, the promise is that the magic will make everything better (and shareholders happier). Despite knowing full-well that AI has an environmental impact of its own: and one that at its current trajectory cannot be kept in line with necessary reduction targets. The AI-fication of everything is presented as a solution — in search of a problem.
