Vitalik Buterin Says AI Data Center Buildout Could Wait Four Years

Buterin Backs Data Center Skeptics

Ethereum ($ETH) co-founder Vitalik Buterin has sided with the movement against large-scale AI data center construction, arguing that algorithmic efficiency, not additional buildings, should deliver the computing gains the industry is planning for. In a post on X published on Sunday, Nov 8, Buterin said he does not agree with the sharp criticism aimed at groups pushing back against the buildout, and set out a case for slowing it deliberately. The position puts him against the direction of capital in the mining sector, where operators are racing to rebrand power assets as AI real estate.

His case is arithmetical. As AI algorithms improve, the same tasks will require fewer computing resources than they do today. Work that would demand larger and more numerous data centers two years from now could, on his timeline, be handled by the infrastructure that already exists by year four, because better software will have done what the new facilities were meant to do. A slower construction pace therefore would not seriously hold the technology back.

He framed the pause as a safety measure rather than an economic one. Delaying the rollout would buy time to prepare for the cybersecurity threats and potential biological risks that come with AI's rapid spread. Scientific progress, he added, does not depend solely on AI compute: a one-year delay in AI development does not push scientific outcomes back by the same period.

His position stops short of hard intervention. Buterin did not call for the immediate closure of existing data centers and did not endorse specific government bans, positioning himself for pacing rather than prohibition. The post also continues a pattern: he has previously tested AI tools built around personal data protection, keeping his engagement on the risk side of the technology he helps steer. That record gives the Sunday post more weight than a one-off remark.

Riot and Cipher Race to Convert

The backdrop is a mining sector converting itself at speed, and the pace of the conversion explains why Buterin's Sunday intervention lands where it does. Riot Platforms signed a 15-year lease on Wednesday, Nov 4 for 910 megawatts (MW) of computing capacity at its Corsicana, Texas facility, with expected revenue of $28 billion from the contract. Including existing agreements, Riot's total contracted data center capacity now stands at 1,151 MW and long-term lease revenue reaches roughly $38 billion. The company paid for the pivot by selling: 20,264 BTC in the first nine months of the year, worth about $1.5 billion, went mainly into data center development. The Corsicana contract alone is worth more than eighteen times the Bitcoin Riot sold in that period.

Cipher Digital is expanding on a parallel track. The company recorded its first high-performance computing lease revenue as rent collection began at its Black Pearl facility, and extended the lease term at Barber Lake to 20 years, lifting expected contract revenue to about $9.1 billion. Projects currently under development total around 5.9 gigawatts. Both operators have, in effect, become landlords for AI compute tenants, with Bitcoin miners now the funding source rather than the business.

Texas is where the two stories collide. Communities have raised concerns over power rates and water scarcity, and Governor Greg Abbott directed in September that permits be paused while the impact of data centers on grid stability and water resources is reviewed. Gina Hinojosa, who won the Texas governor's race on Tuesday, Nov 3, has proposed a stop to new data center approvals until community protections are in place, a move that would, if enacted, freeze the permitting pipeline both companies depend on. The market reacted fast: Cipher Digital shares dropped about 19% the day after the election, although the proposal has not yet been written into state law.

Efficiency Is Buterin's Own Playbook

The argument reads as Buterin's own playbook applied to someone else's industry. He steered Ethereum from energy-hungry mining to proof-of-stake, where staking validators secure the chain at a fraction of the power draw, and pushed scaling through Layer 2 networks rather than raw capacity, a preference his recent record makes explicit: an estimate of 700,000 $ETH in annual subsidies, an argument that Ethereum burns do not guarantee a value floor, and a warning over EIP-8037. Applied to AI, the same logic says compute can be deferred if software improves fast enough. For the companies building 910 MW at a time, the binding constraint is not his opinion but Texas permitting, where an approval freeze is now a live electoral promise and Cipher's 19% drop shows the market pricing it.