The artificial intelligence infrastructure boom is creating a growing conflict between the climate commitments and investment strategies of Microsoft, Amazon, Google, Meta and Oracle. As COP31 negotiations continue in Antalya, Turkiye, the world’s largest technology companies are accelerating spending on AI data centers while electricity consumption, greenhouse gas emissions and dependence on natural gas challenge their clean-energy targets.
The International Energy Agency’s 2026 report, Key Questions on Energy and AI, estimates that capital expenditure by five major technology companies exceeded $400 billion in 2025 and could increase another 75 percent in 2026, implying annual spending of more than $700 billion. The investment covers AI computing infrastructure and other capital assets, not exclusively data centers.
Meanwhile, a November 11 report from Bloomberg, published during COP31, highlights growing concerns about the lack of detailed environmental disclosures for individual AI data centers. CDP, a global environmental disclosure organization, is seeking more granular information on electricity consumption, emissions and water use.
AI Data Center Electricity Demand Could Reach 950 TWh by 2030
The IEA estimates that global data center electricity consumption increased 17 percent in 2025, reaching approximately 485 TWh.
Its 2026 assessment projects consumption of approximately 950 TWh by 2030, with AI-focused data center electricity demand tripling over the period.
Global data center electricity consumption
TWh per year, IEA estimates and projections
0 TWh250 TWh500 TWh750 TWh1K TWh202420252030 forecast
Source: IEA, Energy and AI (2025) and Key Questions on Energy and AI (2026).
The IEA also estimates that approximately 20 percent of planned data center projects could face delays because of grid constraints, transformer shortages, electricity infrastructure limitations and permitting challenges.
Microsoft: Emissions Rise 25% Despite 100% Renewable Electricity Matching
Microsoft’s 2026 environmental disclosures reveal growing pressure on its climate commitments.
The company’s total greenhouse gas emissions have increased approximately 25 percent from its 2020 baseline, driven partly by AI infrastructure construction and supply-chain activities.
Microsoft reported a 24 percent increase in electricity consumption during 2025. Its reported emissions from purchased electricity increased 945 percent compared with 2024, partly reflecting changes in renewable electricity accounting and procurement.
Although Microsoft matched 100 percent of annual electricity consumption with renewable electricity, the company is also developing data centers in Texas and West Virginia that will depend on natural gas generation.
Microsoft has committed to becoming carbon negative by 2030, but its infrastructure expansion is making that target more difficult to achieve.
Amazon: Data Center Expansion Drives 34% Increase in Electricity-Related Emissions
Amazon’s 2025 Sustainability Report shows that indirect emissions from purchased electricity increased 34 percent in 2025.
The company attributes the increase partly to data center electricity consumption, alongside electrification of its delivery network and buildings.
Amazon added more than 1.2 GW of data center capacity during the fourth quarter of 2025 alone, illustrating the scale of AWS infrastructure expansion.
Purchased electricity emissions represented approximately 5 percent of Amazon’s total reported carbon footprint. Consequently, rising electricity demand is only one component of its environmental challenge; construction, manufacturing, transportation and supply-chain emissions must also be considered.
Google: 24/7 Carbon-Free Electricity Target Faces Rising AI Demand
Google has committed to operating on 24/7 carbon-free electricity by 2030, a more demanding objective than purchasing renewable electricity certificates equivalent to annual consumption.
Under 24/7 matching, a data center must have sufficient carbon-free electricity available during each hour of operation, ideally within the same electricity grid.
However, Google’s emissions increased 51 percent from its 2019 baseline, according to figures cited in an April 2026 Reuters investigation into shareholder concerns about hyperscaler electricity and water consumption.
Investors have questioned whether Google’s renewable procurement and clean-energy infrastructure can expand quickly enough to support its AI investments.
Google’s hourly carbon-free energy performance varies by location. A consolidated, verified 2025 hourly matching percentage was not established from the reviewed disclosures, making it inappropriate to equate Google’s annual renewable electricity purchases with continuous carbon-free operation.
Meta: 32.2 GW of Clean Energy Contracts, but Annual Matching Falls Short of 24/7 Supply
Meta has emerged as one of the world’s largest corporate clean-energy purchasers.
According to S&P Global Energy’s November 2026 analysis, Meta had contracted 32.2 GW of clean-energy capacity globally by September 30, including nearly 30 GW in the United States.
During 2026, the company contracted another 10.6 GW of clean-energy capacity.
Meta reports matching 100 percent of its electricity consumption with clean energy annually since 2020. However, it follows an annual portfolio-matching approach rather than guaranteeing 24/7 carbon-free electricity at every data center.
Its clean-energy portfolio therefore represents significant investment in new generation, but the contracted GW cannot be interpreted as electricity available continuously at its AI facilities.
Oracle: AI Infrastructure Expansion Raises Questions About Power Supply
Oracle’s expanding AI cloud infrastructure is increasing its exposure to electricity availability, power procurement costs and data center construction emissions.
Oracle was among the technology companies that signed the US government’s March 2026 Ratepayer Protection Pledge, committing to measures intended to prevent new data center demand from increasing electricity costs for existing customers.
However, comparable Oracle-specific disclosures for data center electricity consumption, 24/7 carbon-free energy performance and renewable PPA capacity were not verified in the reviewed sources.
This reporting gap limits direct comparisons with Microsoft, Google, Amazon and Meta.
AI Data Centers: Climate Performance Comparison
| Company | Verified climate or energy indicator | Principal challenge |
| Microsoft | Emissions +25% from 2020 baseline; electricity consumption +24% in 2025 | Natural gas dependence and embodied emissions |
| Amazon | Purchased electricity emissions +34% in 2025 | Rapid AWS capacity expansion |
| Emissions +51% from 2019 baseline | Achieving 24/7 carbon-free power | |
| Meta | 32.2 GW of clean-energy contracts | Annual matching versus hourly supply |
| Oracle | Comparable electricity and emissions figures not verified | Limited data center-level disclosure |
Sources: Corporate sustainability reports, Reuters, S&P Global Energy and IEA. Different baselines and reporting scopes prevent a direct emissions ranking.
Natural Gas and Coal Remain Important Sources of Data Center Electricity
The IEA’s analysis of electricity supply for AI estimates that the physical electricity mix supplying global data centers in 2024 included approximately 30 percent coal, 27 percent renewables, 26 percent natural gas and 15 percent nuclear power.
The agency projects that renewable generation serving data centers could increase by more than 450 TWh by 2035, while natural gas generation could increase by approximately 175 TWh.
This explains why renewable PPAs do not automatically eliminate fossil-fuel dependence. A data center may purchase enough renewable electricity annually to offset its consumption while continuing to draw gas- or coal-generated electricity during hours when renewable supply is insufficient.
The risk is particularly significant in US data center clusters where utilities are developing additional gas-fired generation and in coal-intensive electricity markets where new AI infrastructure depends on existing grid supply.
COP31: AI Emissions Could Reach 215 Million Tonnes in 2026
An October 2026 analysis by Bricola Group forecasts that combined greenhouse gas emissions from Alphabet, Amazon Web Services, Meta, Microsoft and Oracle could reach 215 million tonnes of CO2 equivalent in 2026, approximately 57 percent higher than the previous year.
After accounting for renewable energy certificates and other contractual instruments, the modeled total falls to approximately 140 million tonnes.
These are independent forecasts, not verified corporate emissions inventories.
The ITU’s Greening Digital Companies 2026 report provides further evidence of the industry’s challenge. Among 200 digital companies assessed, only 25 reported sourcing 100 percent renewable electricity, while just 81 demonstrated comprehensive climate transition plans.
Can Technology Giants Meet Their Climate Targets?
The evidence indicates that AI infrastructure expansion is increasing electricity consumption and emissions faster than some technology companies can reduce their environmental footprint.
Microsoft faces a widening gap between emissions and its carbon-negative ambition. Google must expand hourly carbon-free electricity availability. Amazon faces rising electricity-related emissions, while Meta’s renewable contracts do not establish continuous clean-power operation. Oracle needs more transparent, comparable disclosures.
For COP31, the decisive indicators should be actual data center electricity consumption, hourly carbon-free electricity performance, new renewable generation commissioned, fossil-fuel electricity dependence and independently verified emissions.
Technology companies are financing substantial clean-energy capacity, but their climate commitments will ultimately depend on whether new electricity supply and grid infrastructure can keep pace with AI investment.
BABURAJAN KIZHAKEDATH
