AI can already be more powerful than Bitcoin – and threaten Bitcoin mining.

A new study suggests that the energy required to run an artificial intelligence system may already exceed the energy required to mine bitcoins.

While this may sound like good news for the Bitcoin mining industry, which is constantly under attack for power consumption, the report from the Bitcoin Policy Institute is criticizing AI as a strong competitor for electricity and hardware Bitcoin. .

The sector's deep pockets mean AI companies have the potential to sell miners for the same amount of electricity. With AI offering up to 25 times more revenue per kilowatt hour (kWh) than Bitcoin, some miners are adding AI processors to their data centers or switching from Bitcoin to AI entirely.

BPI researcher Margot Paez told Cointelegraph: “As long as the revenue per megawatt-hour is higher for AI than for Bitcoin, we will see this trend.”

The AI ​​industry is still in its infancy, but the power needs of generative AI models are impressive. According to Goldman Sachs, a single ChatGPT query consumes nearly 10 times the energy of a typical Google search. MIT Technology Review reports that AI image generation can use as much energy as fully charging a smartphone.

Bitcoin mining's high energy use has led to threats of bans in Europe and bans in New York. According to BPI, the annual energy use of US Bitcoin mining facilities is around 121.13 terawatt hours (TWh), and AI will consume between 20 and 125 TWh by 2023 (AI is housed in data centers used for other operations, making exact figures more difficult).

But this year with a large amount in generative AI, the report estimated that in 2024 AI will use 169 TWh and the growth will increase even more than Bitcoin mining, using approximately 240 TWh in 2027 to mining 160 TWh.

Data centers require large amounts of water to maintain efficiency in cooling their machines, which are powered by AI models. According to a study conducted by Shaoli Ren at the University of California Riverside, between five and 50 questions to chatGPT will drink 500 ml of water.

By comparison, Bitcoin mining in the US alone is estimated to require between 93 and 120 gigaliters of water per year, with each transaction said to use enough water to fill a backyard swimming pool (such estimates are controversial due to concerns about their imprecision).

The deep pockets of AI investors put miners at risk

AI computing profit margins are currently much higher than Bitcoin mining. Mining generates revenue of $0.17 to $0.20 per kilowatt, while revenue from Nvidia graphics processing units used for AI can range from $3 to $5, representing a 17-25-fold difference.

So why don't Bitcoin miners reuse their rigs to run AI to make more money?

Bitcoin miner and crypto asset consultant Anibal Garrido told Cointelegraph that making the leap is not that easy, as miners use application-specific integrated circuit (ASIC) machines designed solely to calculate the Pow protocol's hash, which cannot be reused. AI.

But Bitcoin miners need to be updated and replaced regularly – and the tools themselves can adapt. Paez notes that many Bitcoin miners are upgrading their facilities to accommodate GPUs, and at least one company has made the full transition from Bitcoin mining to AI.

Alex de Vries, a data analyst and researcher at the Vrije Universiteit Amsterdam and De Nederlandsche Bank, told Cointelegraph that electricity competition will only get stronger.

“The pockets of AI companies are deeper than the crypto mining industry,” he said. De Vries believes that AI companies may already be “looking at the power contracts of Bitcoin miners.”

The AI ​​industry explained that during the current AI boom, it wants to quickly get power and equipment and cannot wait for construction projects for several years to build new data centers – which means that the threat to mining is real.

Variable versus fixed

Growing AI power consumption could help change the politics surrounding the electricity used for Bitcoin mining, which is relatively well stocked.

Not all energy consumption is equal, and Bitcoin miners are more flexible and can be shut down or recharged to use surplus, waste or cheap electricity.

AI, on the other hand, requires the models to work correctly “99.9% of the time”. This demand often means that they accept whatever energy is available, which can lead to the use of less environmentally friendly or dangerous energy sources.

Energy-intensive plants often use fossil fuels to address unanticipated surges in demand, exacerbating environmental impacts.

The flexibility offered by Bitcoin mining allows miners to make agreements with governments to stop taking power if the grid becomes saturated. Once the grid is stabilized, miners can continue their work, giving the grid more flexibility to maintain balance.

A BPI report found that US bitcoin miners stop operations between 5% and 31% of the time when electricity prices are too high or run by grid operators.

The study, which collected data from eight U.S. mining facilities between July and September 2023, found that these cuts could prevent 13.6 kilotons of CO2 emissions. This reduction is equivalent to removing 2,951 cars from the road.

While AI isn't as volatile as bitcoin mining, Pease said, “the only way you can control their emissions is through direct investment in renewable energy.”

Environment-agnostic with constant and energy incentives for the industry

Another important difference between the two technologies is the space requirements. Bitcoin mining is location-agnostic, with AI requiring low latency to provide very fast responses, requiring data centers to be located near major metropolitan areas.

That means AI data centers must consume whatever energy is available in those particular locations, while bitcoin miners can shift renewable energy facilities to locations with more energy, such as remote locations where hydroelectricity, solar or wind are abundant.

Advocates argue that the existence of Bitcoin mining can support the transition to renewable energy by providing financial stability for new projects with predictable demand in an otherwise low demand period.

The location-agnostic nature of Bitcoin mining also allows for the use of wasted energy. This could include mining with remote hydroelectric power, capturing excess methane emissions, reusing waste heat to turn on heating, or renewable energy from solar and wind sources that could be tied up due to grid shortages.

Can AI be more effective?

Juan Calvo, senior data engineer and general AI engineer at Datatonic, notes the push to use renewable energy for Bitcoin mining and believes AI has a duty to be more sustainable.

Focusing on the importance of ethical and sustainable choices in technological development, we need to assess whether the ability to do something justifies its performance.

The engineer explained that AI developers have various techniques to increase energy efficiency. These include fine-tuning existing models, employing smaller models for specific tasks, and using cloud solutions that can significantly reduce overall energy consumption.

Hardware developments can play an important role. Graphics manufacturers like NVIDIA are at the forefront of developing specialized hardware that improves performance while consuming less power. The integration between more efficient algorithms and advanced hardware will help solve AI's increasing energy demands in a more sustainable way.

However, de Vries points out that the bigger is better variable in generative AI may undermine these efficiency gains. In the year In a study conducted in 2023, the increasing energy footprint of AI models is driven by the incentive to develop ever-larger models, which increases computational resources and energy demand.

“Efforts to further improve these models by increasing the efficiency of the models and reducing their energy costs may be more effective, thereby negating some of the efficiency gains.”

De Vries compared this variable to the efficiency gains in cryptocurrency mining hardware. As mining becomes more efficient, Bitcoin miners earn more easily.

Subscribe

A very engaging read in Blockchain. It is given once a week.

Daniel Ramírez-Escudero is a journalist who has been immersed in crypto since the beginning of 2017 and has many years of experience in the media industry. He is a crypto junky, passionate about geopolitics and interested in the financial, philosophical and technological revolution brought about by Bitcoin.