A recent analysis indicates that the electricity consumption of artificial intelligence (AI) is projected to account for nearly half of the total power used by global data centers by the end of 2025. This suggests that AI’s energy demands are on track to exceed those of Bitcoin mining in the near future.
The study was conducted by Alex de Vries-Gao, a PhD researcher at the Environmental Research Institute of the Vrije Universiteit Amsterdam. De Vries-Gao has previously focused on tracking the electricity consumption and environmental impact of cryptocurrencies. He shared his findings through detailed evaluations accessible to the public.
Currently, AI is estimated to consume up to one-fifth of the total electricity used in data centers worldwide. While precise measurements remain challenging—primarily because major technology firms do not fully disclose energy usage related to AI model training and operations—de Vries-Gao developed projections based on the supply chain of specialized computational chips designed for AI tasks.
Despite ongoing improvements in hardware and algorithmic efficiency, the research highlights that AI’s overall power demand continues to grow rapidly. This trend underscores the need for greater awareness and strategic energy management within the tech industry.
De Vries-Gao initially believed that his work tracking energy-intensive technologies might decline as cryptocurrencies like Ethereum transitioned toward less power-intensive mechanisms. However, the emergence of large-scale AI systems, such as ChatGPT, brought a new high-energy-consumption technology into focus—especially within competitive and fast-expanding markets. Reflecting on this shift, he remarked, “I thought to myself, oh dear, here we go again.”
The analysis emphasizes that the expanding role of AI warrants broader attention to its sustainability footprint. As the industry grows, understanding and mitigating its energy use becomes increasingly critical. Balancing technological progress with environmental responsibility will be a vital challenge moving forward.
Understanding AI’s Growing Energy Use
Artificial intelligence, particularly deep learning and large language models, requires immense computational power. Training sophisticated models involves processing enormous datasets across high-performance servers, which consume significant electricity. This energy demand is compounded by the need for data storage, cooling systems, and continuous model optimization.
Many leading tech companies are investing heavily in AI research and deployment, accelerating both innovation and energy usage. Without transparent reporting or industry-wide standards, however, accurately quantifying total consumption remains difficult.
Efforts are underway to improve AI energy efficiency through better hardware design, optimized algorithms, and the use of renewable energy sources. Nevertheless, the pace of adoption may outstrip these efficiency gains.
Comparing AI and Bitcoin Energy Footprints
Bitcoin mining has long been criticized for its high electricity consumption, often compared to that of entire countries. The mining process involves solving complex mathematical puzzles to validate transactions and secure the network—a resource-intensive activity.
Recent changes in some cryptocurrency systems, including Ethereum’s shift to a proof-of-stake model, have reduced energy use in the digital currency space. In contrast, AI’s appetite for power is climbing steadily as applications multiply across industries.
By late 2025, AI is expected to not only match but surpass the energy demands of Bitcoin mining. This reflects AI’s expansive integration into areas like healthcare, finance, logistics, and consumer technology.
👉 Explore sustainable technology strategies
The Path to Sustainable AI Development
Addressing AI’s rising energy consumption requires a multi-faceted approach. Key strategies include:
- Enhancing computational efficiency through specialized AI chips
- Prioritizing the use of green energy in data centers
- Developing more energy-conscious machine learning algorithms
- Encouraging corporate transparency regarding energy use
Industry leaders and researchers are increasingly aware of these challenges. Collaborative efforts between technologists, policymakers, and environmental scientists can help align AI growth with global sustainability goals.
Frequently Asked Questions
How is AI’s energy consumption measured?
AI energy use is typically estimated based on the power draw of hardware used for training and inference, multiplied by operational time. Since most companies do not release detailed figures, researchers often rely on chip performance data and data center energy reports.
Why is AI’s energy demand growing so quickly?
The expansion of AI applications across numerous sectors—from automated customer service to advanced research—requires more powerful models and larger infrastructure. This scalability directly increases electricity consumption.
Can AI become more energy-efficient?
Yes. Advances in hardware, such as TPUs and GPUs optimized for AI, along with software improvements like model pruning and quantization, are helping reduce energy use per operation. However, total consumption may still rise due to increased adoption.
How does renewable energy fit into AI’s future?
Many companies are committing to powering data centers with renewable sources. This can significantly lower the carbon footprint of AI operations and support sustainable development.
What role can consumers play in reducing AI’s energy impact?
Users can choose services from providers that prioritize renewable energy and advocate for greater corporate responsibility. Supporting efficient AI models and applications also helps.
Will AI’s energy consumption continue to rise after 2025?
While efficiency gains may slow the growth rate, overall energy use is likely to increase as AI becomes more embedded in economic and social activities. Continuous innovation and regulation will be key to managing this trend.