Vitalik Buterin Says Slowing Data Center Expansion Is Overall More Beneficial Than Harmful
On November 8, Ethereum co-founder Vitalik Buterin posted on X that he disagreed with the strong emotional reaction this week to the anti-data center movement. He pointed out that algorithmic progress means tasks that now require larger and more data centers to complete in two years could be completed in four years, or even sooner, with existing data centers. Therefore, even if the anti-data center movement were extremely successful, it would not trigger any doomsday-like catastrophic consequences.
Vitalik said that slowing compute expansion can buy humanity more time to strengthen responses to the cybersecurity issues that have already begun to emerge and to future biosecurity risks. At the same time, scientific applications are constrained by many factors other than AI, so if AI development slows by one year, the slowdown in scientific progress would be far less than one year. Taken together, slowing the ultra-large-scale expansion of compute appears to be a net positive, and this judgment does not need to rely on controversial arguments such as doomsday scenarios or recursive self-improvement (RSI).
He added that slowing data center construction does not mean “depriving us of access,” and other forms of compute restrictions are also among the mildest means in terms of affecting access; care should be taken not to confuse the two.
Original source:
Title: Vitalik Buterin X post Publisher: vitalik.eth (@VitalikButerin) Publication date: November 8, 2026 URL: https://x.com/VitalikButerin/status/2119377918875079163 Source nature: original post by the person involved Verification status: full post text read Supporting content: disagreed with the strong emotional reaction to the anti-data center movement; algorithmic progress means work that would require larger centers in two years could be completed by existing centers in four years; even extreme success would not bring doomsday consequences; slowing expansion could buy time to address cybersecurity and biosecurity risks; scientific progress is constrained by non-AI factors; no need to rely on doomsday theory or RSI arguments.
ABAB AI Insight
Vitalik treats progress in algorithmic efficiency as the core variable, arguing that the compute resources needed for the same task decline over time. This weakens the necessity of continuously building new ultra-large data centers: work that would require more facilities in two years may be completed within four years by existing facilities. This time gap means that even an extreme restriction on new construction would not create a technological cliff, but would only delay the realization of some capabilities.
Slowing construction is described as buying time for adaptation, not as a permanent ban. Cybersecurity issues have already begun to emerge, while biosecurity risks are listed as a possible future category. The extra time could be used to strengthen defenses, rather than waiting until capabilities are fully mature before responding. Scientific applications are constrained by non-compute factors such as experiments, data collection, and verification, so a one-year slowdown in AI capability does not equal a proportionate one-year delay in overall scientific progress.
The argument explicitly excludes reliance on controversial frameworks. Vitalik said that even without adopting assumptions such as doomsday theory or recursive self-improvement, a slowdown in ultra-large-scale compute expansion still appears to be a net positive when viewed through algorithmic progress, the adaptation window for risks, and the structure of scientific bottlenecks. This distinguishes his position from more radical calls for restrictions, while still supporting limits on the pace of construction.
In a later addition, he further distinguished between slowing data center construction and depriving people of access. The former is positioned as a relatively mild measure, and he warned against directly equating restrictions on compute resources with completely cutting off use. The overall argument focuses on the allocation of time and changes in efficiency, rather than an outright denial of AI development itself.
ABAB News · Cognitive Law
Algorithmic progress causes the compute needed for the same task to decline over time; capabilities that would require larger centers in two years may be realized by existing facilities in four years.
Slowing the construction of ultra-large data centers can buy time to address cybersecurity and biosecurity risks, while scientific progress is constrained by non-AI factors and lags less than AI itself.
This judgment is based on efficiency and time windows, without needing to rely on controversial arguments such as doomsday theory or recursive self-improvement, yet still holds that slowing expansion is a net positive.