AI’s Energy Appetite Surpassing Bitcoin’s Consumption: The Urgent Call for Sustainable Practices

As artificial intelligence (AI) continues to evolve and integrate into every corner of our digital lives, a significant concern is emerging: AI’s energy consumption could soon surpass Bitcoin mining. This prediction isn’t speculative hype; it’s grounded in the findings of environmental experts, analysts, and researchers like Alex de Vries, a PhD candidate at Vrije Universiteit Amsterdam, who has been at the forefront of studying the environmental footprint of digital technologies.

The implications are massive. What was once a concern primarily about cryptocurrency’s energy demands has now shifted to the broader AI ecosystem, which is quickly becoming a significant contributor to global energy use. Big tech companies are driving this growth by developing ever-larger AI models, further increasing energy consumption.

Understanding the Energy Appetite of AI

The energy consumed by AI models—especially large-scale ones like OpenAI’s GPT-4, Meta’s LLaMA, and Google’s Gemini—is vast. These models require enormous computational power, primarily provided by data centers running specialized AI chips like GPUs and TPUs, as well as other specialized AI equipment and ai hardware that drive significant energy use. The specialized computer chips used in these AI models, and the chain for specialized computer hardware involved in their production and deployment, are critical factors in the overall energy footprint. This power demand doesn’t just come from training models but also from deploying them across millions of queries and applications every day.

Training a single large language model (LLM) can consume as much electricity as 100 U.S. homes use in an entire year. According to OpenAI, training GPT-3 used several thousand petaflop/s-days of compute. The electricity cost, when aggregated across the entire AI lifecycle (including hardware production, data center operation, and model inference), is staggering. However, determining exactly how much electricity is consumed throughout the lifecycle remains a challenge. It is also difficult to estimate hardware production accurately, as there is a lack of publicly available device details, making it hard to measure the true energy impact of AI systems.

AI vs Bitcoin Mining: A Growing Rivalry in Energy Use

Bitcoin has long been criticized for its massive energy consumption, with the global Bitcoin network using around 140 TWh/year, similar to the energy consumption of a small country. But AI could soon surpass Bitcoin mining in energy demand, as projections based on current trends and available data indicate a rapid increase in AI-related electricity use.

Sector	Estimated Energy Use (TWh/year)	Notable Trends
Bitcoin Mining	140	Stabilizing with energy-efficient ASICs
Global Data Centers	460–500	Rising due to AI workloads
AI Training & Inference	Projected to exceed 300 by 2027	Growing exponentially

Currently, up to a fifth of all the electricity consumed by data centers globally can be attributed to AI workloads, and this share is expected to require even more power as AI adoption accelerates. Some projections estimate that by 2025, AI could account for nearly half—to half—of all the electricity consumed by data centers globally, underscoring the significant environmental impact of AI’s growing energy appetite.

Alex de Vries’ latest research00103-7) suggests that AI may consume up to 134 TWh annually by 2027, rivaling the current demand of Bitcoin mining. What’s more concerning is that AI energy consumption is growing far more rapidly.

Challenges in Measuring AI’s Energy Use

Despite rising concerns, accurately understanding AI’s energy appetite is challenging. Efforts to understand AI’s energy consumption face significant challenges, including data gaps and inconsistent reporting. Several barriers contribute to this:

Lack of Transparency: Many tech companies do not disclose how much energy their AI models use. The issue of tech companies sharing data remains a major obstacle, and there is a growing call for more scrutiny of their sustainability disclosures.
Publicly Unavailable Hardware Specs: Estimating energy usage becomes harder when chip-level data isn’t released.
Variable Data Center Efficiency: Not all data centers are created equal—PUE (Power Usage Effectiveness) ratings can vary significantly.
Inconsistent Reporting: While some companies mention AI energy use in their earnings calls, these companies earnings calls rarely provide quantifiable data, making it difficult to assess the true impact.

“We’re seeing an increase in AI workloads, but companies remain tight-lipped about the energy their AI models consume,” says Alex de Vries, whose Gao-based analysis00103-7/fulltext) outlines these transparency gaps.

Previous research and other researchers have also highlighted the difficulties in measuring AI’s energy consumption, emphasizing the need for better data and methodologies.

The Environmental Impact of AI

The surge in AI development has clear environmental implications. Environmental studies have begun to analyze the impact of AI’s energy use, highlighting the need for more comprehensive data. In regions where clean energy is not prevalent, the energy demands of AI-centric facilities are often met by fossil fuels, increasing their carbon footprint. Unlike traditional data centers, AI-centric facilities require higher compute densities, greater cooling requirements, and specialized chips that are often more energy-intensive. The environmental impact of AI specifically is significant, as its power usage can be difficult to quantify and is distinct from other digital technologies. Comparisons to mining in energy consumption, such as cryptocurrency mining, underscore how both AI and crypto contribute substantially to global power usage and environmental challenges. AI journalism plays a crucial role in promoting transparency and awareness of AI’s environmental impact.

Key Environmental Concerns:

Greenhouse Gas Emissions: If AI workloads are powered by fossil-fuel-based grids, emissions will spike.
Water Usage: Cooling AI data centers consumes millions of gallons of water annually.
Stress on Power Grids: Electricity demand from AI is forcing some regions to delay or rethink data center expansions.

Several studies, including IEA’s 2024 digital energy report, predict that AI-related electricity demand could reach up to 4% of global consumption by 2030.

The Need for Transparent Sustainability Reporting

As AI’s energy consumption becomes a significant part of the tech industry’s carbon footprint, there’s a pressing need for transparent sustainability reporting. A new analysis highlights the growing importance of transparency in AI’s energy consumption, emphasizing how recent findings shed light on the increasing power demand of AI systems.

Recommendations for Tech Companies:

Disclose model-level energy usage
Adopt independent third-party audits
Publish supply chain environmental data
Provide model training and inference energy metrics

Organizations like Partnership on AI are encouraging responsible AI practices, but voluntary compliance is not enough. The push for accountability must be legally and industry-enforced.

Exploring Solutions: Sustainable AI Development

Reducing AI’s energy appetite doesn’t mean halting innovation—it means smarter, more sustainable development. Recent efficiency gains in AI hardware and software have improved energy use, but it remains uncertain whether these improvements are enough to offset the rapid growth in demand. Despite efficiency gains, overall energy consumption from AI continues to rise as larger models and more intensive applications are deployed. To support this growth, organizations must expand data center capacity to accommodate increased power requirements and ensure reliable infrastructure for AI workloads.

Strategies for Sustainable AI:

Model Efficiency Improvements: Use smaller, optimized models where possible (e.g., DistilBERT over BERT).
Sparse Models: Technologies like Mixture of Experts activate only parts of a network to save compute.
Renewable-Powered Data Centers: Shift to solar, wind, or hydro-powered facilities.
Edge AI: Deploying models on local devices reduces cloud compute dependency.

For example, Microsoft and Google are investing heavily in zero-carbon data centers to curb their AI footprint—a step in the right direction but not yet industry-wide.

The Role of Policy and Regulation

Governments and regulators have a vital role to play in managing the AI energy boom. Without intervention, the power demands of AI could overwhelm infrastructure, exacerbate emissions, and contribute to energy inequality.

Possible Regulatory Approaches:

Mandatory Energy Disclosures for AI model training and inference
Carbon Limits for new data centers
Subsidies for Green AI Research
Data Center Permitting Laws tied to energy mix and usage

The European Union’s AI Act and the U.S. Department of Energy’s green computing initiatives are early steps, but global alignment is needed.

Engaging the Tech Community

The tech industry itself must step up. Tech giants like Google and Microsoft are driving rapid AI development, leading to increased energy consumption and a larger carbon footprint. Managing data center capacity is becoming critical to ensure the sustainable growth of AI technologies as rising energy demands put pressure on existing infrastructure. Beyond corporations, developers, researchers, and AI engineers can contribute by:

Using energy-efficient model architectures
Opting for eco-friendly cloud providers
Advocating for open-source energy benchmarks
Supporting initiatives like Green Software Foundation

There’s also a need for education and awareness—ensuring future developers understand the carbon cost of compute and are equipped with tools for sustainable coding.

Final Thoughts: Navigating AI’s Future Responsibly

The rise of artificial intelligence represents one of humanity’s most profound technological shifts—but it comes at a steep energy cost. If AI’s energy consumption surpasses Bitcoin mining, we risk exacerbating climate change, stressing global power grids, and compromising long-term sustainability.

This doesn’t have to be the case. With collective action, clear regulations, and innovation in energy efficiency, we can build an AI-powered future that’s also environmentally conscious.

To learn more about how AI impacts the environment and how you can contribute to more sustainable development, check out our related content:

Stay informed, stay sustainable.

Join our newsletter for weekly updates on AI, sustainability, and cutting-edge tech.

Tags: ai’s energy appetite surpassing bitcoin’s consumption, energy consumption, ai’s energy consumption, soon surpass bitcoin mining, environmental impact, power demand, gao a phd candidate