Six years ago, Google confidently announced that by 2030, all of its global operations would be powered by clean energy sources such as wind and solar, and achieve net-zero carbon emissions. Microsoft also made a bold promise to achieve carbon negative emissions by 2030 – removing more carbon dioxide than it emits.
Nowadays, Google refers to these goals as the “Moonshot Initiative”, suggesting that their realization is as challenging as reaching the moon. Microsoft, on the other hand, redefines its efforts to reduce emissions as “a marathon, not a sprint”.
What has happened to the world in just six years? The answer is “AI”. These tech giants who once loudly proclaimed the slogan of carbon neutrality are now caught in an awkward reality under the influence of the AI boom: on one side, there are ambitious environmental protection commitments; on the other side, there are power-consuming data centers that are astonishingly energy-intensive. The balance is tilting towards the latter.
01 | Carbon Ledger in the Age of AI
Data from the Clean Energy Buyers’ Association shows that in 2024 and 2025, the scale of clean energy purchased by technology companies reached record levels. According to the sustainability reports released by various companies, in the first five years or so after their climate commitments were made, the total emissions actually increased instead of decreasing: Google’s emissions soared by nearly 50%, Amazon’s rose by 33%, Microsoft’s increased by more than 23%, and Meta’s skyrocketed by over 60%.
Behind these figures lies the massive power consumption caused by the training and inference of AI large models. The electricity consumed for training one GPT-3 is approximately equivalent to the annual electricity usage of 130 American households. And the training clusters of the next-generation models have reached a peak power consumption that is close to the output power of a small nuclear power plant.
Patrick Huang, a senior analyst at Wood Mackenzie, said: “Even though these companies haven’t officially revised their targets yet, they are already acknowledging, ‘Yes, we might have strayed from the right path.'”
In the current intense AI competition, technology companies are facing a harsh reality: To remain competitive, they must utilize all available electricity. And in an increasing number of cases, this means using natural gas. The main component of natural gas is methane, which is a greenhouse gas that exacerbates global warming.
According to data from the International Energy Agency, in 2024, the proportion of natural gas in the electricity supply of data centers in the United States exceeded 40%; globally, the proportion of coal reached 30%. This trend shows no sign of abating. Utility companies across the United States are planning to build natural gas power plants to meet the power supply demands of data centers; some technology companies are also planning to build their own dedicated gas-fired power plants specifically for their own data centers.
The technology companies argued that they have made significant progress in reducing emissions through measures such as improving energy efficiency, purchasing green certificates for renewable energy and zero-carbon electricity, and requiring suppliers to reduce emissions. However, the reality is far more complex than what is depicted in the “carbon neutrality” figures.
In the state of Wisconsin in the United States, the two newly-built natural gas power plants that supply power to Microsoft’s data centers will achieve carbon offset through the solar energy projects that the company has invested in other areas of the state. Similarly, the three natural gas power plants will supply power to the Meta large-scale data center in rural areas of Louisiana, and the company will also invest in solar energy projects in other regions. Google also stated that it is investing in wind energy, hydropower, energy storage and advanced nuclear energy, but it also relies on natural gas. The company plans to purchase electricity from a newly-built natural gas power plant in Decatur, Illinois. The plant’s emissions will be captured and stored underground.
However, this “cross-regional offsetting” approach is facing increasing criticism. After the proposed revision of greenhouse gas accounting rules, the power source of the data center must be in the same region as the data center itself, and the power supply time must also match. For instance, solar green certificates can only be used to offset electricity consumption during the daytime operation period.
Lori Bird, the director of the US energy program at the World Resources Institute, described it as: “Companies are frantically competing for electricity resources, striving to obtain as much electricity as possible as quickly as possible. This is a fierce battle for resources.”
02 | The Dilemma of Tech Giants
Facing the energy abyss brought about by AI, the tech giants are shifting from being “model students” to “realists”, and are now directly intervening in the energy infrastructure.
In January 2026, Meta announced three nuclear energy agreements, collaborating with Vistra, TerraPOWER and Oklo, aiming to support up to 6.6 gigawatts of new and existing nuclear power generation capacity by 2035. This includes its partnership with TerraPOWER to support the development of two Natrium sodium-cooled fast reactor units, its collaboration with Oklo to build an “advanced nuclear technology park” in Ohio, and its cooperation with Vistra to support the expansion and upgrade of three nuclear power plants. Meta stated that this is the “largest-scale support to advanced nuclear technology to date” by the company.
Microsoft has gone even further – it has restarted the Three Mile Island nuclear power plant to provide stable power for AI data centers. Google and Amazon are also investing in small modular reactors (SMRs). In addition, the tech giants are also turning their attention to nuclear fusion. Sam Altman, the CEO of OpenAI, is one of the main proponents of nuclear fusion. He believes that new energy sources need to be developed immediately to support the growth trajectory of the technology industry.
The path of emission reduction for tech giants is still confronted with severe challenges at the policy level. Since President Trump took office last year, he has systematically suppressed renewable energy. He cancelled the funding and licensing for solar and wind power projects, as well as the tax incentives for renewable energy. Environmentalists claim that these projects are less costly and faster to build compared to natural gas or nuclear power plants. At the same time, Trump ordered that several coal-fired power plants that were planned to be decommissioned should continue to operate. He also claimed that climate change was a “hoax”, arguing that green energy was unreliable, costly, and could even undermine the United States’ energy independence.
In March 2026, three Democratic senators from the United States initiated an investigation into eight AI companies, covering 12 data center projects of companies such as xAI, OpenAI and Meta. These projects will be powered by large-scale newly-built gas power plants, and many of these plants have a size more than ten times that of the average existing gas power plants in the US grid. The senators warned in the letter that these power plants would emit hundreds of millions of tons of greenhouse gases and tens of thousands of tons of harmful pollutants each year. “If there are not sufficient measures to reduce methane emissions, the carbon intensity of gas-fired power generation could be comparable to that of coal-fired power generation.” If all the gas projects currently under development in the United States were completed, they would cumulatively emit 1.21 billion tons of carbon dioxide over their lifespan, which is twice the annual emissions from all other sources in the United States.
03 | The clock of the glaciers will not stop
Julie McNamara, the deputy policy director of the Climate and Energy Program of the Union of Concerned Scientists, believes that the sharp increase in electricity consumption of data centers has “completely turned the emission reduction challenge into a crisis”.
In this AI-driven energy revolution, Chinese enterprises are playing an increasingly important role. On March 31, 2026, Yang Rui, the chairman of Shuang Deng Co., Ltd., made a judgment at the company’s annual new product launch event: “In the future, intelligent computing data centers (AIDC) will no longer merely be major electricity consumers, but will be a highly coupled ‘computing power + energy’ composite system. Under this backdrop, the role of energy storage will undergo a fundamental change, becoming a ‘standard configuration’ for computing infrastructure and even a decisive force.”
This assertion is based on the insight that the essence of AI competition is shifting from an algorithmic battle to an energy battle. As AI progresses from generative to inferential, computing power moves from kilo-node clusters to ten-thousand-node and one-hundred-thousand-node clusters, and data centers are evolving from information infrastructure to a computing power industrial system. This transformation has brought about a brand-new concept: power computing synergy. “Whoever can make the most efficient use of electricity will be able to make the best use of computing power; whoever can build a superior energy system will be able to support the next generation of computing infrastructure.”
Yang Rui’s judgment is being increasingly validated by facts. Nidea Technology is a prime example. Morgan Stanley believes that Nidea Technology has the system integration capabilities to address the pain points at the grid level. Its abundant funds provide a guarantee for its continuous expansion and technological investment. The research report predicts that Nidea Technology’s annual compound growth rate of profits from 2026 to 2027 will reach 22.8%, and it is expected to maintain an annualized growth rate of over 20% until 2030.
In the fields of solid-state batteries and sodium batteries, Chinese enterprises are also at the forefront. Goxuan High-tech has reached a strategic partnership with Enjie Co., Ltd., and CATL and Changan Automobile jointly launched the world’s first sodium-ion production passenger vehicle. These new technologies are expected to further reduce the cost of energy storage and enhance safety in the future. It is worth noting that China’s policy planning in the field of green fuels is also accelerating. In March 2026, a special symposium of the National Energy Administration clearly stated that green fuels are an important direction for ensuring energy security, promoting green development, and facilitating the non-electric utilization of new energy, and they are a key driver for new quality productive forces in the energy sector.
On the other side of the Earth, in the Andes Mountains of Peru, scientists are recording the latest data on the retreat of glaciers. Over the past century, the Andes Mountains have lost more than half of their glaciers. These slowly formed ice masses are not only a reserve of fresh water but also a repository of geological time. The United Nations Intergovernmental Panel on Climate Change (IPCC) has repeatedly warned that if global carbon emissions are not reduced by half by 2030, the goal of limiting the temperature increase to 1.5℃ will be completely unattainable. Every single snowflake contributes to the formation of glaciers, and each choice shapes the future climate landscape.