Recently, Mistral AI provided a rare benchmark in its industry environmental disclosure, detailing the footprint of its flagship large-scale language model Mistral Large 2. The report shows that training and running the model generated 20.4 kilotonnes of carbon dioxide equivalent emissions, consumed 281,000 cubic meters of water, and consumed 660 kilograms of antimony equivalent materials in the 18 months ending in January 2025. It is worth noting that a 400-token reply from its chatbot Le Chat only used 1.14 grams of carbon dioxide, 45 milliliters of water, and 0.16 milligrams of mineral resources.
During this year's World Earth Day, PowerVerse also released the energy and environmental impacts of AI large language models and Bitcoin blockchain network mining. A peer-reviewed paper published in the scientific journal Joule pointed out that by the end of 2025, AI's electricity consumption may account for 49% of global data center electricity usage, exceeding Bitcoin's well-known "energy appetite."
How does AI's energy consumption compare to Bitcoin's carbon footprint? Bitcoin's energy usage has been an ongoing topic of debate and is often used as a reason to ban Bitcoin mining. This makes AI reasoning particularly frugal relative to Bitcoin's proof-of-work engine.
On average, a single Bitcoin transaction emits 600 to 700 kilograms of CO2, consumes more than 17,000 litres of water, and produces more than 130 grams of electronic waste. According to the Cambridge Centre for Alternative Finance, the entire Bitcoin network emitted approximately 48 million tonnes of CO2 in 2023, consumed more than 2 billion litres of water, and produced more than 20,000 tonnes of electronic waste. However, while the Cambridge figures are peer-reviewed, they have also been subject to considerable criticism and come with important caveats.
First, Bitcoin’s electricity mix isn’t singular. According to a miner survey conducted by BTC investment fund Batcoinz as of March 2023, hydropower (23.1%), wind (13.9%), and solar (5%) together account for more than 40% of Bitcoin’s energy consumption. The difference is because Batcoinz’s survey included off-grid power generation. Nuclear power, which is generally considered carbon-neutral, accounts for another 7.9%. Natural gas and coal together account for 44%, but Bitcoin’s energy mix is more diverse than critics assume.
Second, large language models (LLMs) may benefit from cleaner grids by default , and the distribution of grid infrastructure affects energy consumption in different areas . For example, nuclear power accounts for more than 22% of the total electricity generation in the European Union, which reduces the CO2 emissions associated with model training and inference in EU data centers such as Mistral. If trained in coal-rich areas in the United States, it will present a completely different environmental impact.
The importance of infrastructure
The energy consumption caused by AI large language models and Bitcoin transactions all operate in an infrastructure environment, either using clean energy and renewable energy, or still using non-renewable energy such as coal resources.
Infrastructure greatly shapes their environmental impact. Training cutting-edge models like GPT-4 or Gemini can still require millions of GPU hours and significant water consumption, while Bitcoin is designed to be mined every 10 minutes regardless of demand, resulting in a fixed energy consumption that increases over time rather than usage.
On the one hand, the strong demand for energy has forced technology giants to take action to update the way they obtain energy. On the other hand, we have also seen that DePIN is playing an increasingly important role.
On June 3 this year , Meta signed a 20-year agreement with Constellation Energy to purchase nuclear power, marking Meta's first official entry into the nuclear energy field to meet the growing power needs of data centers.
Google pledged to fund the development of three new nuclear power projects this year, and previously worked with Kairos Power, a developer of small modular reactors (SMRs). Amazon also invested more than $500 million in the development of SMRs last October, and acquired a data center park powered by the Susquehanna Nuclear Power Plant in March 2024. In March of this year, Amazon, Google and Meta also jointly signed an initiative initiated by the World Nuclear Association, calling for a tripling of global nuclear power capacity by 2050.
In addition to updating the way energy is obtained, energy can also be fully utilized through innovative ways in terms of energy use. DePIN has become another way to supplement energy consumption by obtaining benefits through blockchain, distributed nodes, and resource sharing.
Taking the AI language model, which has huge computing power requirements, as an example, training a large model may take days or even weeks and consume a lot of electricity. It is estimated that the energy consumed by the GPT-3 training process is equivalent to the electricity used by a household for several years. This high energy consumption not only has a negative impact on the environment, but also increases operating costs and brings sustainability issues.
Decentralized cloud computing in the DePIN field connects idle computing resources to the cloud computing blockchain network, provides high-performance, low-latency, and low-cost computing resources to computing power demanders, and obtains profits by "on-chaining idle computing resources". While making full use of computing resources, it allows computing resources to flow fully and participate in the market mechanism, avoiding waste and solving the computing power bottleneck problem.
The halving mechanism of the Bitcoin blockchain steadily reduces the rate at which new coins are generated, incentivizing miners to become more efficient over time , and its energy consumption is consistent with the practicality of protecting the decentralized global financial network. The public and transparent data on energy consumption of Mistral AI large-scale model training also allows the public to understand its environmental impact.
Continuing improvements in clean energy adoption , innovative ways to use energy , and mining optimization will be key to the expansion of Bitcoin and AI as core pillars of the digital economy.
refer to:
https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai
Disclaimer: This article is an original article from Power Verse and does not constitute investment and application advice. You are welcome to share this article with more friends. When media, brands and related platforms reprint it, they must indicate the source as Power Verse .
