Rising data center emissions are exposing weak points in Big Tech’s climate strategies.
In its April 2025 report “Energy and AI”, the International Energy Agency (IEA) stated: “While the share of data centers in aggregate emissions may appear small, data centers are among the few sectors – along with road transport and aviation – projected to experience an increase in both direct and indirect emissions through 2030. In the Lift-Off Case, data centers exhibit the largest emissions growth of any sector.”
Given this outlook for rising energy demand and emissions—particularly in a scenario of accelerating AI adoption—we took a deeper look at the decarbonisation strategies of eight of the largest data center operators with a global presence, including several of the Big Tech firms.
We find that most of the companies are pulling many of the critical decarbonisation levers — but some important ones are still being overlooked. In view of rising emissions, companies could increase the credibility of their climate strategies by taking additional measures.
What are the levers for reducing data center emissions?
Data centers generate emissions in several ways: through the construction of their facilities and the purchase of hardware—both classified as Scope 3 emissions—and through the electricity required to operate that hardware, which falls under Scope 2 emissions.
Among the eight companies we analysed, Scope 2 and Scope 3 emissions have increased significantly—by an average of 8–11% per year between 2020 to 2023—coinciding with the rapid adoption of generative AI and cloud computing.
For some companies, particularly when new data centers are built, the Scope 3 emissions from data center construction surpassed the Scope 2 emissions from electricity use. This underscores the importance of considering Scope 3 emissions in efforts to decarbonize data centers.
We identified ten effective levers across four categories to reduce — or at least minimize — these emissions going forward (see Table 1).
The four categories were:
- Renewable energy procurement: Securing clean electricity is the most effective way to reduce a data center’s operational emissions. Impactful sourcing strategies include direct procurement through Power Purchase Agreements (PPAs) and the use of high-quality Renewable Energy Certificates (RECs), both of which provide long-term revenue certainty to new renewable energy projects and support grid decarbonisation.
In contrast, low-impact approaches — such as purchasing unbundled, low-quality RECs from existing projects or acquiring certificates on the spot market — may offer little to no additional environmental benefit. These strategies often fail to drive new renewable capacity and can undermine the credibility of climate targets.
In addition, companies can engage with electricity suppliers through green tariff programs or strategically site data centers in regions with low grid carbon intensity or natural advantages like ambient cooling, which reduce electricity demand. - Energy efficiency measures: Energy efficiency is another critical lever. Measures range from using more efficient computer chips, power savings, deploying more efficient cooling systems and AI-driven optimisation of energy use to upgrading server hardware and shifting workloads to times or locations with greater renewable energy availability. Advanced strategies include carbon-aware computing, where operations are scheduled based on real-time grid emissions data.
- Levers for low-carbon construction: To address embodied carbon — emissions associated with construction materials and infrastructure — companies can adopt low-carbon concrete, recycled steel, and sustainable materials. Modular, prefabricated, and design-for-disassembly approaches can also reduce emissions, as can repurposing existing buildings where feasible.
- IT equipment circularity: Finally, IT hardware represents a significant source of Scope 3 emissions due to manufacturing, transport, and disposal impacts. Extending hardware lifespans, enabling reuse, and implementing circular practices in procurement and end-of-life management can help mitigate these emissions and build resilience in supply chains.
What emissions reduction levers are data center companies disclosing?
Building on this framework, we analyzed the latest sustainability disclosures1 of major technology firms that operate data centers — Microsoft, Amazon (AWS), Google, and Meta — alongside large colocation data center providers Equinix, Digital Realty, NTT Data, and CyrusOne. Collectively, these companies operate over 1,000 data centers, accounting for approximately 16% of all data centers worldwide.
Colocation providers offer third-party facilities where multiple customers can rent space, power, and cooling infrastructure to host their servers, making them critical enablers of the broader digital economy.
Our assessment evaluated the extent to which each company reports using the key decarbonisation levers described above in their climate strategies and disclosed actions. Each lever was scored based on the strength of the available evidence, using the following scale:
- 0 = no evidence that the lever is considered;
- 1 = partial evidence, where the lever is mentioned but its strategic relevance is unclear;
- 2 = strong evidence that the lever is strategically important to the company’s climate approach.
We found that while most companies disclose key decarbonisation measures – such as procuring renewable energy, improving cooling systems, and, to a lesser extent, using sustainable building materials – they have yet to systematically apply the full range of available solutions.
Renewable energy: procurement is the key focus area
All companies engage in impactful procurement of renewable energy. Six companies provide strong evidence that they prioritize high-impact procurement, primarily through the use of long-term power purchase agreements (PPAs), where available. Notably, Google stands out with its more complex strategy to match local renewable energy demand on an hourly basis rather than on an annual scale.
While we found limited evidence in Amazon’s sustainability disclosures, it’s important to highlight that—alongside Microsoft, Google and Meta—they have recently become the largest purchasers of clean energy in the U.S. Much of the recent surge in corporate green power procurement has been driven by Big Tech.
In this context, Google has acknowledged that its data center energy demand is outpacing its ability to source clean electricity in an impactful way, particularly outside of Europe—signaling a potential future challenge as data center growth continues.
It is also worth noting that Amazon, Microsoft and Google increasingly frame their strategies around “carbon-free energy,” a broader category that includes nuclear power. This may indicate a future shift away from pure renewables in order to meet the scale and reliability demands of their data center operations.
In contrast to their clean energy procurement efforts, we found limited evidence of companies proactively engaging with electric utilities—aside from Amazon and Meta—or prioritising data center siting based on grid carbon intensity.
In fact, the concentration of data centers in traditional regions such as Northern Virginia and Texas appears to be increasing. That said, there are early signs of a shift in location strategies, particularly in response to the infrastructure demands of AI. Some AI workloads can be placed in non-traditional locations, potentially improving their environmental performance. For example, CyrusOne has indicated that its AI workload-specific centers could prioritize sites based on sustainability criteria such as access to renewable energy, low water stress, and moderate temperatures—rather than relying solely on proximity to major network or population centers.
Energy efficiency: time to leverage all technology solutions
Energy efficiency is a common area of focus, particularly in cooling system improvements and water-saving technologies. Six companies disclose strong efforts in this area.
In our analysis, we found that Microsoft’s disclosures from the previous year did not provide sufficient evidence of a strategic focus on implementing more efficient cooling solutions. However, recent disclosures indicate that the company has undertaken substantial research into the trade-offs associated with various cooling technologies. Their latest CSR report highlights a strategic emphasis and a technological focus on a specific cooling solution, demonstrating progress in this regard.
Interestingly, colocation providers are more likely than AI-focused tech companies to use AI for energy optimisation—a notable gap given the latter’s technological capabilities.
We also found that colocation firms Equinix, Digital Realty, and CyrusOne have set firm-wide targets for Power Usage Effectiveness (PUE) or Carbon Usage Effectiveness (CUE)[2]—key efficiency metrics for data centers that are also mandatory to disclose in the EU —reflecting a strategic focus on environmental performance. In contrast, we found no evidence of comparable targets among the major tech companies.
Among large technology firms, improving server efficiency—such as through more energy-efficient chips—is somewhat more prevalent than implementing carbon-aware compute scheduling. These practices are not applicable to colocation companies, which do not operate the servers themselves.
Low-carbon construction: A blind spot for some companies
Only half of the companies reviewed provide strong evidence of efforts to reduce embodied carbon in construction, mainly through the use of low-carbon building materials. Tech giants tend to disclose more robust action in this area than colocation providers. However, practices such as modular construction, prefabrication, or infrastructure reuse are rarely reported across the board.
Reusing and recycling IT equipment: A common practice
All companies analyzed show strong efforts to enable hardware reuse, although only Microsoft discloses a quantified target for reusing and recycling servers and components across its entire cloud hardware fleet. Again, circularity measures are not applicable to colocation providers, who do not own or manage the hardware.
Overall, CyrusOne showed the most comprehensive application of decarbonisation measures among colocation providers. Among the Big Tech companies, Microsoft stood out as a leader.
Conclusion
As data center emissions continue to rise despite current efforts, companies would do well to broaden their current approach. Demonstrating the credibility of future climate strategies means going beyond renewable energy — and that starts with using the full toolkit. This includes deploying all available decarbonisation levers, especially tech-driven solutions they can develop and control in-house.
For investors, this is a key moment to dig deeper. Climate risk analysis shouldn’t stop at headline renewable energy targets, which may become increasingly difficult to achieve. Instead, it should focus on whether companies are taking detailed, measurable action across the entire value chain — from energy efficiency and embodied carbon to circularity and site selection.
References
- Sustainability reports and public CDP disclosures, where available; accessed 5 June 2025
- Power Usage Effectiveness (PUE) is a metric used to determine the energy efficiency of a data center, calculated as the ratio of total facility energy consumption to the energy used by the IT equipment alone. A lower PUE indicates a more energy-efficient data center. Carbon Usage Effectiveness (CUE) is a metric that measures the carbon emissions associated with data center operations, defined as the ratio of total carbon emissions from the data center to the energy consumption of the IT equipment. It provides insight into the environmental impact of the data center’s energy usage.
Author Information
