Vitalik Proposes Crypto Solutions to Cut AI Energy Use

Key Highlights:

• Sanders calls for a pause on building massive AI data centres to cut energy use and emissions.
• Buterin favours a “pause button” and decentralization to control mega-clusters without stopping AI innovation.
• Crypto and decentralized AI could grow together which can offer safer, energy-efficient alternatives to giant data centres.

Bernie Sanders is proposing a temporary halt on building massive data centres that power AI systems, arguing that their energy consumption is putting a great amount of strain on power grids, which increase emissions, and prioritizing tech giants’ needs over public energy access.

These hyperscale facilities, run by companies such as Microsoft, Google and Amazon, consume power comparable to small nations, mainly to support rapidly growing AI models such as GPT and Claude.

Sanders believes that a pause would force the policymakers and companies to reassess the environmental, social and geopolitical costs of unchecked AI expansion.

To this proposal, Ethereum co-founder Vitalik Buterin disagrees and states that a slowdown alone is not the right solution. He responded to the tweet by an X user (NathanpmYoung) and argued that AI development will continue even after the pause and simply pausing infrastructure growth fails to address the root issue, which is thinking about how AI is built and powered.

In his post, he urges people to rethink AI infrastructure, focusing on energy-efficient models, cleaner power sources, and alternative architectures that reduce reliance on centralized, energy-hungry data centres.

The Case for a ‘Computing Power Pause’

Vitalik in his post is suggesting a different solution than Sanders. Instead of just slowing down the construction of AI data centres, he proposes creating a “pause button” for AI systems. This mechanism could cut the available computing power by 90-99% within a year or two in case of emergencies, such as a rogue AI or military misuse.

Buterin is comparing his AI “pause button” idea to circuit breakers in blockchain networks. Just as circuit breakers temporarily stop a blockchain to prevent problems, the pause button could temporarily cut AI computing power in emergencies without having to stop normal AI development.

Preference for Slowdown Plus Decentralization

Buterin personally supports a slower, more decentralized path for AI development. According to him, we should treat huge, power-hungry data centres differently from regular AI devices like personal GPUs or small edge computers.

By making rules for only the big centres, we can stop a few companies from controlling everything and let researchers as well as smaller projects keep working.

Crypto Parallels and Market Opportunities

Blockchain supporters can see parallels between AI and crypto industries. For example, Ethereum cut its energy use a lot by switching from proof-of-work to proof-of-stake. Some projects such as Bittensor, and AI tools on Solana, use idle GPUs from around the world and do not rely on huge centralized data centres.

Moreover, if the U.S. introduces new rules, it could push crypto and AI to work together more safely and efficiently.

Risks of Loopholes and Evasion

However, Buterin has also warned that simply trying to limit AI computing power could be easy to bypass. Companies might split up big data centres, move them to countries with lenient rules like parts of Southeast Asia, or even turn regular consumer hardware into hidden supercomputers.

Events such as chip smuggling to China, show how hard it actually is to enforce restrictions. Cutting computing power by 90-99% would need strict global monitoring, maybe even treaties similar to those for nuclear weapons.

Whatever Buterin is suggesting offers a smarter middle path than just stopping the construction. If Sander’s proposal passes, it could push AI toward smaller, more distributed systems and let crypto networks play a bigger role in making AI safer and more energy-efficient.

Also Read: Ethereum Completes Fusaka Upgrade with Zero Downtime