By the end of 2025, artificial intelligence may surpass Bitcoin mining in electricity consumption.
AI accounts for up to 20% of data center energy consumption, which is growing rapidly.
Despite the optimization of algorithms and increased equipment efficiency, the power requirements of the AI industry are growing exponentially.
This situation is reminiscent of the cryptocurrency boom of 2017-2021, when mining Bitcoin and other digital assets led to a sharp increase in electricity demand.
A similar scenario is unfolding in artificial intelligence, and the consequences could be even greater, affecting not only cryptocurrencies, but also leading futures such as Dow Jones index futures and S&P 500 futures.Â
The AI industry follows the same logic as cryptocurrency mining: the more power the model requires, the more efficient it is.
Companies like OpenAI, Google, and Meta constantly increase the parameters of neural networks, which consume substantial energy resources.
For example, GPT-4 training required dozens of times more energy than GPT-3.
This trend drives a boom in the construction of data centers, especially in the United States, where about 40% of the world’s data centers are already concentrated.
Operators are planning new gas-fired power plants and even modular nuclear reactors to meet their energy demands.
However, such solutions strain the grid and jeopardize climate goals, as some of the energy generation remains carbon-intensive.
The primary similarity is high power consumption because AI requires enormous computing power, like mining.
In addition, the concentration in regions with cheap energy shows that AI data centers, like mining farms, are being built in Texas, Canada, and Scandinavia, with access to affordable electricity.
The pressure on the power grid is no less critical because sudden spikes in demand can lead to overloads, as has already happened in Iran and Kazakhstan due to mining.
Another similarity between AI and cryptocurrencies is the difficulty in estimating real-world energy consumption.
Large companies like Google and Nvidia publish reports on carbon dioxide emissions, but do not specify AI’s contribution.
For example, it’s known that Google’s carbon footprint has grown by 13% in recent years, but how much of this is due to AI models remains a mystery.
Similar problems exist in the crypto industry: mining pools rarely disclose accurate consumption data, and estimates vary significantly, from 70 to 150 TWh per year for Bitcoin only.
How does this affect the crypto market and the price of Bitcoin?
- Competition for energy leads to rising electricity prices, resulting in lower mining profitability.
- If AI takes over the lion’s share of the capacity, miners will either have to move to new regions or close down.
- Some data centers in the United States are already displacing mining farms due to the higher margins of AI.
- The authorities may impose taxes on energy-intensive AI and cryptocurrency industries.
- There may be restrictions on constructing new data centers in energy-deficient regions.
These all lead to stricter regulations and hash rate — a metric representing computational powers the Bitcoin network uses — growth will slow down.
If mining becomes less profitable, some of the equipment will shut down, and the network’s complexity will be adjusted, but the Bitcoin price may receive support due to lower sales by miners and automated trading software used in arbitrage.
Artificial intelligence follows the path of cryptocurrencies, but on a much larger scale.
If the trend continues, by 2025, the AI industry will exceed Bitcoin mining in energy consumption, creating new challenges for energy systems and the environment.
This means tougher competition for resources and possible pressure on miners for the crypto market.
However, if Bitcoin retains the status of digital gold, its price may be less dependent on energy consumption than the cost of computing for AI.
One thing may soon become true: the era of cheap energy for high-tech industries is ending.
Back to Daily Market News.
About The Author
