AI Has A Climate Problem — And A Climate Opportunity

Climate Week in New York is making one thing increasingly clear: the future of artificial intelligence and the future of energy are becoming impossible to discuss separately. As Climate Week NYC runs from September 20 through September 27, 2026, more than 1,000 events are bringing together conversations around energy, electrification, technology, artificial intelligence, finance and the transition to a lower-carbon economy. One of the week’s most timely conversations is happening around the enormous infrastructure required to keep AI running — and whether that same technology could eventually help build a smarter, cleaner energy system.

The timing matters because AI is no longer simply a software story. Behind every large language model, image generator, AI assistant and automated system is a physical network of data centers, processors, cooling equipment, electricity infrastructure and communications systems. As AI adoption accelerates, the amount of computing infrastructure required to support it is growing as well. Climate Week’s official AI Forum on September 22 is addressing precisely this tension, describing AI demand as something that is outpacing infrastructure, energy supply and natural resources while also examining whether AI could help accelerate renewable-energy investment and climate action.

That creates a fascinating contradiction. AI needs more electricity, but AI could also become one of the technologies used to manage electricity more intelligently. The same technology placing new demands on the grid could potentially help utilities forecast demand, integrate renewable energy, manage batteries, improve building efficiency and operate increasingly complicated energy systems. The climate question surrounding AI, therefore, is not simply whether artificial intelligence consumes too much energy. It is also about what society does with the intelligence that becomes available as the technology develops.

The electricity demand is already becoming measurable. The U.S. Energy Information Administration expects American electricity consumption to reach record levels in both 2026 and 2027, with AI-focused data centers and broader electrification among the factors driving the increase. Reuters reported earlier this month that U.S. electricity consumption is projected to rise from 4,195 billion kilowatt-hours in 2025 to 4,270 billion in 2026 and 4,349 billion in 2027.

That growth puts the data center at the center of a much larger infrastructure conversation. A data center may look like a technology facility, but it is also effectively a large energy customer. It needs electricity, cooling, networking, backup systems and reliable access to infrastructure. When dozens or hundreds of facilities are planned in the same region, the question becomes whether the existing grid can support that additional load without creating bottlenecks or shifting costs onto other electricity users.

This is why the energy source behind AI matters so much. If additional computing demand is supplied by electricity generated from lower-carbon sources, the climate impact can be different from a system that relies heavily on fossil-fuel generation. Renewable energy therefore enters the AI conversation not as a separate environmental issue, but as part of the infrastructure needed to support the next generation of computing.

Solar and wind power bring their own challenge, however: they are variable sources of electricity. Solar generation changes throughout the day, while wind production depends on weather conditions. AI and other advanced computing systems, meanwhile, may require highly reliable electricity. Connecting those two worlds requires better forecasting, storage, transmission and grid management.

That is where artificial intelligence could become part of the opportunity. AI systems can analyze enormous amounts of information and identify patterns that may be difficult to manage using traditional approaches alone. Energy companies can use advanced software and machine learning for tasks such as forecasting electricity demand, predicting renewable-energy output, identifying equipment problems and optimizing operations.

The technology does not magically create electricity, but it can potentially help people make better decisions about how electricity is generated, stored, distributed and consumed. That distinction is important. AI is unlikely to be a single solution to climate change, but it could become one of the tools used to operate a more complicated energy system.

Climate Week NYC is already treating that relationship as an integrated problem. A September 24 working roundtable organized by the Cornell Atkinson Center for Sustainability is explicitly asking how the electricity grid can become cleaner while meeting the growing power demand created by AI and data centers. Rather than treating decarbonization and rising electricity demand as completely separate challenges, the event frames them as problems that need to be solved together.

That may be the most important shift in the conversation.

For years, discussions about climate technology and artificial intelligence often happened in completely different rooms. Climate discussions focused on renewable energy, emissions, electric vehicles, batteries and efficiency. Technology discussions focused on processors, software, cloud computing and AI models. Now the two industries are increasingly dependent on each other.

The AI industry needs power. The energy industry needs better tools for managing an increasingly complicated grid. Renewable-energy developers need infrastructure that can move electricity from generation sites to customers. Data-center operators need reliable power. Utilities need better forecasts. Businesses want more efficient buildings. Consumers are adopting more electric devices. All of these trends are happening simultaneously.

The result is a much larger infrastructure puzzle.

One particularly important piece is the physical grid itself. Electricity generation is only useful if that electricity can reach the places where it is needed. Transmission lines, substations, transformers and other infrastructure can become bottlenecks even when additional generation capacity exists elsewhere. That means the AI race is increasingly connected to an infrastructure race.

Reuters recently described electricity as a critical factor in the next phase of industrial competition, pointing to the growing importance of cheap, scalable and cleaner electricity as digital infrastructure, electrification and decarbonization increase demand. AI and data centers are part of that broader transformation.

Cooling is another part of the equation. Data centers generate significant amounts of heat, and keeping computing equipment operating requires sophisticated cooling systems. Depending on the technology and location, cooling can involve substantial energy and water use. That means the environmental footprint of an AI facility cannot be measured only by looking at the electricity consumed by its processors.

The location of infrastructure matters too. A data center built in a region with abundant low-carbon electricity, strong transmission capacity and appropriate cooling resources faces a different set of environmental circumstances from a facility built in an area already dealing with electricity or water constraints.

This is why sustainable AI is becoming an infrastructure question rather than simply a software question.

The industry is also looking at efficiency. More efficient processors can perform more computing work using less electricity. Better cooling systems can reduce the energy required to remove heat. Improved software can reduce unnecessary computation. Data-center operators can optimize workloads and infrastructure. These changes can make individual AI tasks less energy-intensive.

But there is an important complication: efficiency can encourage greater use.

If AI becomes cheaper and easier to run, people may use more of it. Businesses may deploy AI across more operations. Consumers may interact with AI more frequently. Companies may build larger models and more applications. The result can be a situation where the energy required for each individual computation falls while total computing demand continues to rise.

That is why the climate conversation cannot stop at efficiency.

The bigger question is how quickly overall AI demand is growing compared with improvements in efficiency and the expansion of cleaner electricity.

There is also a significant opportunity on the renewable-energy side. AI could help renewable-energy operators forecast production more accurately, allowing grid managers to anticipate changes in solar and wind generation. It could help determine when batteries should charge or discharge. It could identify patterns in building energy consumption and recommend ways to reduce waste. It could help detect maintenance problems before equipment fails.

These may sound like relatively small applications compared with generative AI, but they could become increasingly important as electricity systems become more decentralized and complicated.

The energy system of the future may contain millions of connected devices. Electric vehicles could charge at different times. Home batteries could store electricity and feed it back into the grid. Rooftop solar systems could generate power locally. Large renewable-energy projects could feed electricity into transmission networks. Buildings could automatically adjust their energy use. Data centers could become major electricity customers with sophisticated demand-management systems.

Managing all of those moving pieces will require enormous amounts of information.

AI could potentially help coordinate them.

That does not mean the technology should be treated as an automatic climate solution. The benefits depend on how AI is designed and deployed, what energy sources support it, how infrastructure is regulated, and whether investment in generation and transmission keeps pace with demand.

The risks are already becoming part of public discussions. Reuters recently reported on growing concerns about the effects of data-center development on electricity demand, infrastructure and local communities in parts of the United States. Other recent reporting has highlighted the increasing competition for power and the infrastructure challenges associated with the rapid expansion of AI facilities.

That makes Climate Week’s AI conversation especially relevant. The issue is no longer simply what AI can do on a computer screen. It is about what AI requires in the physical world.

Every new model has an infrastructure story.

Every data center has an electricity story.

Every electricity source has an environmental story.

And every new demand on the grid creates questions about how that system should expand.

At the same time, renewable energy itself is continuing to grow. Recent Reuters reporting noted that renewables are expected to account for an increasing share of global electricity generation, while renewable energy and battery storage represent the majority of planned new U.S. generation capacity over the coming year.

That creates the possibility of a feedback loop in which the growth of AI encourages greater investment in energy infrastructure, while improvements in renewable energy and grid technology make it easier to support additional computing.

The opportunity is not simply to build more power plants for data centers.

It is to build a more flexible and intelligent energy system overall.

That could mean better transmission. It could mean more battery storage. It could mean smarter demand response. It could mean improved renewable-energy forecasting. It could mean more efficient buildings and industrial facilities. It could mean data centers becoming more sophisticated participants in electricity markets rather than simply large consumers of power.

The challenge is that all of this infrastructure takes time.

AI development can move incredibly quickly. A new model can be trained, released and adopted within months. Building a transmission line, upgrading a substation or developing a major renewable-energy project can take years. That difference in speed is one of the fundamental challenges facing the AI-energy relationship.

Technology can move at software speed while infrastructure often moves at construction speed.

Climate Week is highlighting that gap because closing it will require cooperation across industries that traditionally operated separately. Technology companies, utilities, energy developers, investors, regulators, engineers and climate organizations increasingly have reasons to work together.

The question is therefore becoming less about whether AI is a climate problem or a climate opportunity.

It can be both.

AI creates additional energy demand, but it can also provide tools for managing energy demand. Data centers consume electricity, but their growth can also stimulate investment in generation, storage and grid infrastructure. AI requires resources, but the technology may help improve the efficiency of other resource-intensive systems.

The outcome will depend on what gets built around the technology.

If AI infrastructure expands without enough attention to electricity generation, transmission, water, cooling and local impacts, its environmental challenges will become harder to ignore. If computing growth is matched with investment in cleaner electricity, storage, grid modernization and efficiency, AI could become part of a broader infrastructure transformation.

That is the bigger story emerging from Climate Week NYC.

The future of artificial intelligence is not happening entirely inside laboratories or on computer screens. It is happening in power plants, renewable-energy projects, transmission corridors, data centers, battery facilities and the buildings that consume electricity every day.

The future of climate technology is changing too. It is increasingly about using software, sensors, automation and data to make physical systems smarter.

AI may have a climate problem.

But it also has a climate opportunity.

The real challenge is making sure the opportunity grows faster than the problem.