Live·Open questions in longevity research
All news
Scientific ComputingIndustry & policy

Vitalik Buterin supports slowing construction of AI data centers

9 November 2026· 261109003

Vitalik Buterin supports slowing construction of AI data centers

On November 8, the co-founder of the Ethereum cryptocurrency network pushed back against harsh criticism of those opposing new data centers. He expects improvements in algorithms to partly offset slower growth in computing capacity, while the additional time would allow stronger defenses against cyberattacks and biological threats.

In Buterin’s scenario, a task that would require more and larger data centers in two years could be handled by today’s data centers in four years, probably sooner. More efficient algorithms should deliver the required result with less computation. In a separate reply, he distinguishes slowing construction from denying people access to AI.

In researcher Adam Marblestone’s model, more efficient software could accelerate AI development itself. Existing computers could run more copies of AI systems that search for and test new algorithms. Successful improvements would then free up more computing resources for research.

The pace of scientific research using AI also depends on many other conditions, Buterin adds. If research faces other obstacles, a delay in AI could partly overlap with the time scientists would spend overcoming those obstacles anyway. This leads to his prediction for scientific research:

“Slowing AI by one year will delay research by much less than a year.”

Buterin proposes using the additional time to protect computer systems and prepare for biological threats. In his January 2025 essay, he had already proposed prioritizing defensive technologies, such as tools for detecting dangerous infections early and checking software security.

Buterin has himself funded efforts to combat aging. In February 2018, the SENS Research Foundation, which was developing ways to repair age-related damage, reported receiving a donation from him of 2,4 million dollars in cryptocurrency. Now life extension advocate Linus Petersson asks him:

“Setting aside the risk of human extinction, what if this delays the point at which aging is overcome?”

As the debate continues, Petersson argues that developing a treatment for aging is much harder than distributing it afterward. He considers government regulation one of the necessary conditions for development itself. If a therapy takes longer to arrive, some people who could have benefited from it will die while waiting.

AI researcher Ben Goertzel agrees that more efficient algorithms are beneficial, but considers slower construction harmful to humanity overall. He points to data centers that have already been approved or are under construction and argues that restrictions in the US would also weaken its position in global competition. Researchers need computing capacity to develop more efficient algorithms too, he says.

Originally published on Telegram by Ukhvat NewsView on Telegram
Sources
#ai-data-centers#compute-scaling#algorithmic-efficiency#ai-safety#aging-research