Artificial intelligence is transforming the global technology industry, but its rapid expansion is bringing a growing environmental challenge into focus: the enormous amount of electricity and computing power required to operate advanced AI systems.

From conversational assistants and image generators to sophisticated coding tools, AI applications depend on powerful data centres packed with specialised processors. These facilities require continuous electricity to operate and additional resources to keep their equipment cool.

As companies such as OpenAI, Anthropic and DeepSeek compete to develop increasingly capable models, questions are emerging about how much energy these technologies consume and what their growing popularity means for global carbon emissions.

The Hidden Energy Behind Artificial Intelligence

Every interaction with an AI assistant requires computing resources. When someone submits a question, the request is processed by specialised hardware inside a data centre before a response is generated.

A simple text question may require relatively little electricity. However, more demanding activities, including generating images, analysing large documents, writing complex software and performing multi-step reasoning, can require substantially more computing power.

The environmental impact also extends beyond individual requests. AI companies must operate large computing facilities around the clock, maintain cooling infrastructure and regularly upgrade their hardware.

According to estimates published by Our World in Data, data centres consumed approximately 1.5% of global electricity in 2025. AI-related computing is expected to become a major driver of additional demand in the coming years.

OpenAI, Claude and DeepSeek: Who Uses More Energy?

Comparing the environmental performance of leading AI companies is not straightforward.

OpenAI, Anthropic and DeepSeek use different models, computing architectures and infrastructure arrangements. Their energy requirements also depend on the complexity of individual tasks, the length of responses and the hardware used to generate them.

OpenAI’s ChatGPT supports a wide range of activities, from everyday conversations to advanced reasoning and programming. More demanding requests can require significantly more computation than basic interactions.

Anthropic’s Claude is widely used for document analysis, software development and extended reasoning. These activities can involve lengthy processing sessions that consume more resources than ordinary chatbot conversations.

DeepSeek has attracted international attention for its competitive AI models and alternative approaches to model development. However, lower development or operating costs do not automatically mean lower electricity consumption.

Independent researchers are increasingly developing tools to estimate the energy requirements of models whose companies do not publish detailed operational figures.

One such initiative, the CLEER dataset, developed by the Sustainable AI Group, provides estimated energy consumption ranges for selected closed AI models by comparing them with experimentally evaluated open models.

These estimates offer useful insights, but they should not be treated as official electricity disclosures from the companies themselves.

Why Complex AI Tasks Consume More Electricity

Not all AI requests have the same environmental footprint.

A short question about general knowledge requires considerably less processing than asking an AI system to develop an application, analyse extensive financial information or complete a complicated research assignment.

AI agents can make the difference even more significant. Unlike conventional chatbots that may generate a single response, agent-based systems can perform multiple steps, use external tools, review information and repeatedly process results.

This extended activity increases the amount of computation required.

Research highlighted in 2026 has demonstrated that complex AI tasks can have environmental footprints dramatically larger than straightforward questions.

Consequently, measuring AI energy consumption through a single average figure can create a misleading picture of the technology’s actual environmental impact.

Data Centres Are Becoming a Major Energy Challenge

The rapid growth of AI has accelerated investment in large-scale data centres across the United States, Europe and Asia.

These facilities require reliable electricity supplies because even short interruptions can affect thousands of servers and critical digital services.

Cooling is another important consideration. High-performance processors generate substantial heat, requiring sophisticated air-conditioning or liquid-cooling systems.

In regions where electricity production relies heavily on coal and natural gas, increasing data-centre demand can contribute to additional greenhouse-gas emissions.

The pressure is not limited to electricity generation. Data centres can also place demands on local water resources, power infrastructure and land availability.

As AI adoption increases, governments and energy companies are facing difficult questions about how to provide sufficient electricity without undermining climate targets.

Can More Efficient AI Models Reduce Emissions?

Despite these concerns, technological improvements could help reduce the environmental cost of artificial intelligence.

Developers are exploring several approaches to make AI systems more efficient:

• Smaller AI models: Compact models can perform specific tasks with fewer computing resources.
• Advanced processors: More efficient chips can deliver greater computing performance for each unit of electricity.
• Renewable energy: Solar, wind and other low-carbon electricity sources can reduce emissions associated with data-centre operations.
• Smarter cooling: Improved cooling systems can reduce the additional energy required to maintain data-centre temperatures.
• Efficient AI processing: Optimising software and reducing unnecessary computational steps can lower energy requirements.

Another important development is the growing interest in measuring energy consumption more accurately.

Without consistent reporting standards, businesses and consumers cannot easily determine which AI systems are genuinely more environmentally efficient.

The Transparency Problem in the AI Industry

One of the biggest obstacles to understanding AI’s environmental footprint is the limited availability of detailed information.

Major technology companies generally do not disclose comprehensive, model-specific electricity consumption figures covering every type of AI request.

The actual footprint depends on several factors, including processor efficiency, data-centre location, electricity sources, cooling technology and workload intensity.

This makes direct comparisons between ChatGPT, Claude and DeepSeek particularly challenging.

Independent benchmarks can help fill some of these information gaps, but their findings depend on assumptions and estimation methods.

Greater transparency could allow regulators, investors and customers to evaluate the environmental performance of AI products more effectively.

What Does This Mean for the Future of AI?

The competition to develop more powerful artificial intelligence is unlikely to slow down. Businesses, governments and consumers are increasingly integrating AI into everyday operations.

However, the next stage of technological development will require more than improvements in intelligence, speed and accuracy.

Energy efficiency and environmental responsibility are likely to become increasingly important measures of technological success.

Companies that can deliver powerful AI services while reducing electricity consumption and carbon emissions may gain an important advantage.

The challenge will be to ensure that the benefits of artificial intelligence do not come at an unnecessarily high environmental cost.

Conclusion: AI’s Future Depends on More Than Computing Power

Artificial intelligence has the potential to improve healthcare, education, scientific research, manufacturing and countless other industries.

But its rapid expansion also creates a growing demand for electricity, infrastructure and natural resources.

The environmental impact of OpenAI, Claude, DeepSeek and other leading AI platforms cannot yet be measured through one universally accepted figure. More transparent reporting and independent testing will be essential to understanding the true costs.

Ultimately, the future of AI will not be determined solely by which company develops the smartest model. It will also depend on which companies can make advanced intelligence more efficient, accessible and environmentally sustainable.