AI vs Crypto Mining Clashes Over U.S. Energy Resources

As the growth of artificial intelligence devours an increasing share of the electricity supply, leading AI vendors like Amazon are vying with bitcoin miners for limited power resources. The AI vs crypto miner competition is part of a long-term trend in which power-hungry data centers are forecast to massively increase their power and water consumption, threatening major environmental impact.

KEY TAKEAWAYS

•Major tech companies are working to reduce their environmental impact, and more than 100 data center operators and industry organizations have committed to an agreement to make data centers climate neutral by 2030.(Jump to Section)
•This trend is reshaping the crypto industry, as some bitcoin miners retrofit their facilities to serve AI vendors. (Jump to Section)
•
The AI vs crypto energy competition is driven by the ever-rising energy consumption of data centers, which are forecast to double their percentage of US electricity use by 2030. (Jump to Section)

AI vs Crypto: David vs Goliath

The battle for energy resources between the crypto miners and AI vendors is akin to David vs Goliath, with the AI providers clearly the Goliath due to their vast balance sheets. The March sale of a Pennsylvania nuclear-powered data center owned by Talen Energy demonstrated the imbalance. With a market cap of $4.48 billion, Marathon Digital Holdings was among the rumored potential purchasers. However, the company that finally won the sale was Amazon Web Services, which benefits from Amazon’s gargantuan $1.83 trillion market cap. AWS paid $650 million for the 960-megawatt data center.

“Large scale bitcoin miners that have been building utility-scale sites are competing head-on with the AI companies” for power resources, said Fred Thiel, CEO of bitcoin mining company Marathon Digital Holdings. 

In an interview on CNBC, Thiel said bitcoin miners are more flexible in their power use than the AI vendors. “We have the ability to shut off any time the grid needs it… so when energy needs are high in the afternoon, we can shut down our systems,” he said. In contrast, “AI loads are not load-balancers—they just consume.”

The nonstop power consumption of AI vendors is due to the compute-intensive nature of AI, which requires high and constant levels of electricity. For instance, powering the computation of OpenAI’s ChatGPT application requires the vast resources of Microsoft’s Azure cloud platform. And now with an entire generation of AI apps launching—from Midjourney to Claude to countless enterprise AI apps—this same pattern of energy consumption is being repeated again and again.

Disrupting the Bitcoin Miners

The scramble for energy resources is causing a significant disruption among cryptominers. In some cases, smaller bitcoin miners that can’t compete with the mega-cap AI players will be unable to power their energy-hungry businesses. In other cases, bitcoin miners are turning a problem into a revenue source by leasing or selling their fully-powered facilities to AI vendors.

Bit Digital, Hive Blockchain Technologies, Hut 8 Corp, and Core Scientific are among the bitcoin miners who have retrofitted their facilities to serve AI providers. These upgrades require significant expense. Cryptominers must replace ASIC servers specially geared for bitcoin mining with the highly expensive GPU-powered machines needed to support AI applications. Yet the massive investment flowing into AI means that this cost is well covered by the new revenue that bitcoin miners are generating.

Sustainability Concerns

There are approximately 11,000 data centers across the globe, with about a third of them in the U.S. The ever-increasing power demands of these facilities raise major concerns about their environmental impact. Whether the expanding global network of data centers can be powered on a sustainable basis remains an open question.

The Electric Power Research Institute forecasts that by 2030, U.S.-based data centers will consume up to 9 percent of the total electricity generated in the U.S. This forecast represents a greater than 100 percent increase from current levels. 

Similarly, Goldman Sachs forecasts that data center power demand will surge 165 percent by 2030 and that consumption of power will increase from its current 1 to 2 percent of global supply to 3 or 4 percent by the decade’s end.

To cool all their energy-intensive machines, data centers consume enough water to earn a spot on the list of top 10 water-consuming industries in the U.S. Data from the International Energy Agency indicates that data centers generate 3.5 percent of the global greenhouse gas (GHG) emissions, which intensifies global climate change.  

To counter this, many major tech companies are working to reduce their environmental impact. Google has pledged to operate carbon-free by 2030. Microsoft’s goal is to be carbon-negative by 2030, meaning that it plans to emit less carbon than it removes. Internationally, more than 100 data center operators and industry organizations have committed to the European Green Deal, which is an agreement to make data centers climate neutral by 2030.

Learn more about ensuring the responsible use of technology in eWeek’s guide to AI policy and governance.
The post AI vs Crypto Mining Clashes Over U.S. Energy Resources appeared first on eWEEK.