Vitalik Buterin alongside AI data center servers and Ethereum graphics, illustrating debate over slowing AI infrastructure expansion.

Vitalik Buterin Backs Slower AI Data Center Expansion

Key Insights

  • Algorithmic efficiency could soften future demand, but its pace remains uncertain.
  • Long-term leases raise the financial stakes of grid access and project approvals. 
  • Community protections and permit timelines can change the economics of new facilities.

Vitalik Buterin argued on Nov. 8 that opposition to new AI data centers could buy time for safeguards. Those safeguards include stronger cybersecurity and protections against biological threats. Better algorithms, he said, may let existing facilities handle future AI workloads. Those workloads might otherwise require substantially more computing capacity. His argument comes as Bitcoin miners sign multibillion-dollar leases to turn power-rich sites into AI infrastructure.

Why a slower buildout may not stall AI

On X, Buterin rejected harsh criticism aimed at anti-data-center groups during the past week. He said algorithmic gains could change how quickly developers need new capacity. Work that may need larger centers in two years could run on today’s facilities within four.

That forecast supports his claim that even major success for opponents would not trigger apocalyptic consequences. However, Buterin did not recommend shutting down operating centers or a ban. Rather, he said, slower construction would allow engineers and public agencies to work out the known security shortcomings.

He also wondered if there was a direct impact of using more AI compute on scientific discovery. Experiments, data gathering, review and deployment cause bottlenecks other than computing. So, a one year delay in learning by AI might be less than a year delay in some science. He didn’t say that he didn’t see the potential for AI to speed up research. Rather, he wondered if “compute is the sole factor in the speed of discovery.” Even as processing power increases, results can be limited by laboratory access and funding, data quality, and data validation.

Vitalik Buterin stated that society has to wait for a while to build up resistance against cyber-attacks and potential biological threats. He argued that without relying on any prediction of the inevitable AI catastrophe.

Permits and leases reshape the debate

The policy debate predates the post. On Sept. 21, Governor Greg Abbott directed state environmental regulators to pause data center permits. The effects on power reliability and water resources would be reviewed by regulators.

Vitalik Buterin in October conducted tests of privacy-centric AI, aiming to minimize data exposure. Those experiments addressed user privacy rather than construction policy.

Cipher Digital reported first high-performance-computing lease revenue on Nov. 2, after rent began at Black Pearl. It extended its Barber Lake lease to 20 years. The company expects about $9.1 billion in contracted revenue there. It also reported a development pipeline of roughly 5.9 gigawatts.

Gina Hinojosa beat incumbent Darrell Castle to the gubernatorial office on Nov. 3, which included her proposing the pause on new approvals. She urged the legislators to provide safeguards to the energy prices and water sources, as well as public involvement. Her proposal has not become statewide law. Any implementation would depend on actions after she takes office.

On Nov. 4, Riot Platforms disclosed a 15-year, 910-megawatt lease for critical computing at its Corsicana, Texas, site. Riot estimated $28 billion in revenue over the initial term. Including other leases, the company said contracted data center capacity reached 1,151 megawatts. Riot valued that total at about $38 billion in long-term revenue.

Riot also sold 20,264 Bitcoin during the first nine months of 2026, raising about $1.5 billion. It said the proceeds would primarily fund data center development. Cipher shares fell nearly 19% on Nov. 4 as investors assessed the prospect of tighter development rules.

Large power connections and predictable utility costs also shape where mining firms build. An uncertain approval schedule can therefore change financing decisions before construction begins.

The next phase depends on safeguards

It is unclear if software efficiency will compensate for new computing needs. Buterin says that if the algorithmic gains come through for him, then the slower rate of construction could be cheaper than any of his competitors think. Without higher efficiency, capacity for developers may be even more limited as demand increases.

A balance needs to be struck between expansion and safeguards, reliable power and community protections by policymakers and operators. There will be tangible costs to that debate, and assumptions about future algorithms. Both timelines call into question by Vitalik Buterin’s argument. It doesn’t show that stopping will be safe. However, it results in a portion of the expense of waiting being part of the policy cost.

Scroll to Top