Microsoft’s deal to restart a nuclear reactor at Pennsylvania’s Three Mile Island—infamous for its partial meltdown in 1979—underscores the tremendous and rapidly growing demand for power within the tech sector. If the facility comes back online, it will produce enough energy to power roughly 800,000 homes per year—but that would go toward Microsoft data centers.
The tech industry currently consumes around 2% of global electricity demand. The EIA estimates that those figures could double by 2026, thanks to growing demand from data centers, artificial intelligence, and cryptocurrency mining. These three sources alone could consume more than 800TWh of electricity each year over the next few years.
In this article, we’ll look at how these dynamics affect different markets and what might be in store for the future.
How Much Power Does Crypto Use?
Cryptocurrencies like Bitcoin that use proof-of-work algorithms require significant computational power to mint new coins—and it requires more computation effort every year to achieve the same outcome. These high costs drive the scarcity that supports their prices in many ways. Currently, the energy to mine just one Bitcoin could power 61 U.S. homes for a year!
Not surprisingly, the demand for electricity to power cryptocurrency mining operations has grown significantly over the past several years. According to the EIA, cryptocurrencies consume around 110TWh of electricity annually, accounting for roughly 0.4% of global annual electric demand—or as much as the Netherlands’ total energy consumption.
The EIA expects these figures to rise by 40% to 160TWh by 2026—but with a critical caveat. Ethereum, the second largest cryptocurrency, reduced its electricity demand by an astounding 99% in 2022 by changing to a proof-of-stake mechanism. If Bitcoin made a similar change or became less popular, these power requirements could move sharply lower.
Crypto mining is also very elastic. If energy prices rise in the afternoon, miners can shut down and resume when energy prices start falling again. These dynamics could limit the impact of crypto mining on the energy prices consumers pay. And it could help pick up any excess capacity that would otherwise go to waste as renewables become widespread.
How Much Power Does AI Use?
Artificial intelligence—and large language models (LLMs), in particular—consume enormous computation resources. Since they rely on discrete mathematics, these models require resource-hungry GPUs rather than CPUs. A single ChatGPT query uses 2.9Wh of electricity compared to just 0.3Wh for a Google search!
Of course, the demand for electricity to power LLMs and other machine-learning models has also risen in recent years. Based on an EIA assessment using NVIDIA GPU unit sales, the AI industry consumes about 7.3 TWh of electricity per year. Most of that comes from companies like OpenAI and Anthropic.
While that’s less than crypto’s 110 TWh of energy consumption, projections estimate that AI energy will grow much faster, consuming at least ten times as much energy between 2023 and 2026, according to the latest estimates. Many new competitors are entering the market, and companies requiring their models will need more computational power.
Unlike crypto miners, AI typically involves non-stop power consumption, limiting their ability to take advantage of off-peak hours. As a result, these increases could have more of an impact on everyday electricity consumers as they compete for electricity. This has led to more widespread efforts among AI companies to acquire their own power sources—like Microsoft.
Impact on the Energy Markets
The impact of rising demand for electricity depends on the region. For instance, Nordic countries benefit from lower electricity costs since the colder weather translates to lower cooling demand. Countries like the U.S. and China also have much lower wholesale electricity costs compared to the European Union. These dynamics influence the location companies choose to base.
Projections suggest that the electricity demand from crypto will grow between 2022 and 2026, but AI energy usage will go from nearly nothing to a significant piece of the pie. Source: IEA
The good news is that both AI and crypto are progressing in reducing energy intensity.
Once we have a trained large language mode, much less electricity is needed to run the model and generate responses. These dynamics are paving the way for private, on-device LLMs that could reduce the usage of online options. Moreover, hardware manufacturers are developing specialized processors that could reduce power consumption.
In the crypto space, many cryptocurrencies are transitioning to a proof-of-stake model following Ethereum’s move. By eliminating proof-of-work, these changes virtually eliminate electricity requirements and could be a game-changer. But, of course, Bitcoin alone comprises roughly half the market and shows no sign of switching away.
Renewable energy could also play a growing role in both markets. While crypto mining has a poor track record (about 65% of mining occurs in China where coal makes up 60% of the energy mix), renewables have become increasingly popular over time. And many data centers preferred by AI companies have embraced nuclear, solar, and other alternatives.
Impact on the Crypto Markets
The impact on the crypto market could be more pronounced. With AI competing for computational resources, many smaller crypto miners may find they can no longer compete for energy and hardware. Others are leasing or selling their crypto mining facilities to AI vendors looking to ramp up their training without building their own facilities from scratch.
Bit Digital, Hive Blockchain Technologies, Hut 8 Corp, and Core Scientific are just a handful of Bitcoin miners retrofitting their facilities to assist AI providers, according to eWeek. These upgrades involve swapping ASIC servers with GPU-powered machines to support AI applications, but the massive investment in AI makes the move economical.
If a significant number of miners begin transitioning away from crypto, the market could see less new supply from mining activity, and prices could head higher. We could see these effects amplified by higher energy costs that could make crypto mining less profitable, compounding the number of players abandoning crypto mining and looking toward AI markets.
Ultimately, we could see the impact limited to proof-of-work cryptocurrencies like Bitcoin and not proof-of-stake alternatives like Ethereum.
The Bottom Line
The battle for energy between crypto and AI is reshaping the electricity markets. While crypto currently leads in energy usage, AI’s projected demand will snowball, potentially overtaking crypto before 2026. This surge could impact electricity prices and availability around the globe, with effects varying by region.
Both industries strive for greater energy efficiency, with crypto exploring less energy-intensive validation methods and AI developing specialized hardware and on-device models. The increasing energy demands are also driving industry adaptations, with some crypto-mining operations pivoting to serve AI computational needs.
If you trade crypto assets, ZenLedger can help organize everything for tax season. Our platform aggregates transactions across wallets and exchanges, computes your capital gains or losses, and generates the forms you must file each year. You can even spot opportunities to reduce your year-end tax burden with tax loss harvesting!
This material has been prepared for informational purposes only and should not be interpreted as professional advice. Please seek independent legal, financial, tax, or other advice specific to your particular situation.
