Buterin Backs Slower AI Data-Center Buildout as Texas Debates Limits
Ethereum co-founder Vitalik Buterin says a slower race to build AI data centers could give society more time to prepare for the technology’s risks. His argument rests partly on a possibility, not a certainty: better algorithms may let existing computing facilities do more with less.
- Buterin backs slower expansion, not shutting existing data centers.
- Efficiency gains could reduce computing needs, but total demand may still rise.
- Bitcoin miners Riot Platforms and Cipher Digital are pursuing long-term data-center deals.
- Texas proposals have put electricity, water and community protections under scrutiny.
Buterin’s case for slowing the buildout
In a post dated November 8, 2026, Buterin defended opposition to large data-center projects. He argued that improvements to algorithms, the software methods used to perform computing tasks, could reduce the infrastructure needed for future AI workloads.
He offered a hypothetical comparison: a capability that might require much larger facilities in two years could potentially be achieved with today’s infrastructure in four years, or sooner. That example illustrates possible efficiency gains. It is not a proven forecast of how much capacity future AI will need.
“A slowdown in compute hyper-scaling seems quite net good, ” Buterin wrote.
“Compute hyper-scaling” means rapidly expanding computing capacity, often by building large data centers. Buterin worries that rapid expansion could leave governments, companies and researchers too little time to strengthen defenses against cybersecurity threats and possible biological risks linked to increasingly capable AI.
He also argued that slowing AI development by a year would not necessarily delay scientific advances by the same amount. Research depends on more than computing power, he said. Buterin did not specify what safeguards a slower buildout would make possible or how effective they would be.
His position concerns the pace of expansion, not shutting down facilities already in operation. He did not propose a specific government ban.
Efficiency does not guarantee lower total demand
More efficient algorithms can reduce the computing resources needed for an individual task. But lower costs or faster performance can also encourage wider use and new applications. If that happens, total demand for computing could still grow, even as each task becomes more efficient.
The available information does not show which force will dominate or when. Buterin’s argument makes a case for caution, but it does not prove that planned data-center capacity is unnecessary. Slower construction could create time to address risks, while also constraining infrastructure sought by businesses and researchers.
Bitcoin miners pursue data-center revenue
Bitcoin mining companies are seeking revenue beyond mining by developing or leasing sites for high-performance computing (HPC), which handles demanding workloads such as AI. Their potential advantage lies in land, power access and development infrastructure, not in simply converting Bitcoin mining machines into AI servers. Mining ASICs are built for Bitcoin’s hashing process and cannot be swapped for the processors used in AI workloads.
Riot Platforms said on November 4, 2026, that it signed a 15-year lease for 910 megawatts of what it calls “critical computing capacity” at its Corsicana facility in Texas. Riot estimated the lease would generate $28 billion in contracted revenue over its initial term. Including other leases, the company said it had 1, 151 megawatts of contracted data-center capacity and approximately $38 billion in long-term contracted revenue.
A megawatt measures power capacity. It does not show how much electricity a facility will use over a year or how many computing tasks it can perform. Riot’s announcement uses the term “critical computing capacity” but does not define it in the information provided here.
These revenue figures are amounts expected under agreements over time, not cash already collected, profit or a guarantee that every project will be completed as planned. Riot also said it sold 20, 264 bitcoin during the first nine months of 2026, raising approximately $1.5 billion, primarily to fund data-center development.
In its November 2, 2026, earnings release, Cipher Digital reported that it received its first HPC lease revenue after rent payments began at its Black Pearl site. The company said it extended its Barber Lake lease to 20 years, expected approximately $9.1 billion in contracted revenue from the site and had a development pipeline of roughly 5.9 gigawatts.
Revenue received after rent begins is different from expected payments across a long lease. A development pipeline is a collection of potential projects, not necessarily signed leases or operating capacity. Those distinctions matter when multibillion-dollar figures make an expansion sound more immediate than it is.
Texas weighs data-center impacts and restrictions
Large data centers can raise local concerns about electricity supply, water use, land and permitting. Those concerns deserve scrutiny, but forecasts and public concerns do not prove that a particular project has already raised bills or depleted water supplies.
On September 21, 2026, Texas Governor Greg Abbott directed state environmental regulators to pause data-center permit approvals while officials reviewed potential effects on electricity reliability and water resources. It was part of a review, not a permanent statewide prohibition.
Gina Hinojosa, who won the November 3, 2026, gubernatorial election, proposed suspending new data-center approvals until lawmakers establish protections for communities, including safeguards concerning energy costs, water supplies and public participation. Her proposal had not become statewide law, and its eventual scope remained unsettled.
The policy debate is not simply a choice between economic development and community protections. Data-center projects can bring investment, but they also place demands on infrastructure and local resources. Any restrictions would need to clarify how officials assess those costs and benefits, and how affected communities take part in decisions.
Texas’ review of data-center approvals drew attention as investors weighed the election result and the prospect of tighter development restrictions. Cipher shares fell nearly 19% on November 4, 2026. The timing does not prove that the election alone caused the decline.
Key questions about AI data centers
-
Does Buterin want existing data centers shut down?
No. He argued for slower expansion but did not call for an immediate shutdown or propose a specific government ban.
-
Why does Buterin support a slower buildout?
He argues that it could give society more time to strengthen defenses against cybersecurity threats and possible biological risks. He did not identify specific safeguards or show how a slower buildout would deliver them.
-
Would a one-year AI slowdown delay scientific progress by a year?
Not necessarily, in Buterin’s view. He argues that research depends on more than AI computing power. That is his position, not a measured guarantee.
-
Has Texas enacted Hinojosa’s proposed moratorium?
No. Hinojosa proposed pausing new approvals until lawmakers establish community protections, but the proposal had not become statewide law.
-
Are Riot’s and Cipher’s multibillion-dollar figures revenue already received?
No. The figures describe contracted or expected payments over time. Cipher’s development pipeline represents potential projects, not completed, operating capacity.
The question is how fast to build
Buterin challenges the assumption that faster data-center construction automatically means faster scientific progress. Algorithmic improvements may reduce computing needs for some tasks, but they do not guarantee that total demand will fall. Slower construction may buy time, but only if that time is used to address real risks.
For Bitcoin miners, the shift into data centers reflects the value of power access and suitable sites. Their revenue estimates are substantial, but depend on projects being built, leases being fulfilled and policy conditions remaining workable.
The choice is not between accepting every data center and rejecting them all. The question is whether people can assess the benefits, costs and risks openly, and whether communities get a meaningful say before construction begins.
