§1 — The paper
In May 2018, a Dutch economist named Alex de Vries published a four-page commentary in the journal Joule. It was titled “Bitcoin’s Growing Energy Problem.” It was careful, specific, and hedged. De Vries estimated that Bitcoin’s mining network consumed at least 2.55 gigawatts of electricity — roughly equivalent to Ireland’s total power use — and could peak at 7.67 gigawatts. He noted that each transaction may consume around 300 kilowatt-hours, though this was probably an underestimate. He did not claim Bitcoin was fossil-fuelled. He did not project that its energy use would grow without limit. He flagged significant uncertainty throughout.
That paper has now been cited more than 646 times in peer-reviewed academic literature. It is the single most commonly invoked empirical authority in the Bitcoin energy debate. The question we set out to answer — using an AI classifier to read all 646 papers — is whether what the world took from that paper is what de Vries actually wrote.
The answer is no. Not even close.
§2 — What the literature inherited
We built a classifier that read each of the 646 papers against 16 specific propositions drawn from de Vries’ original text. The results document a pattern that plays out across years and across hundreds of papers with striking consistency.
The hedges disappeared almost immediately. De Vries wrote that Bitcoin’s per-transaction energy footprint may be 300 kilowatt-hours — and flagged this was probably an underestimate. Of the papers that repeated this specific figure, 92.3% stated it as a hard fact. No qualifier. No uncertainty. A sentence de Vries never wrote, now circulating under his authority.
Claims he never made began to accumulate under his name. De Vries’ paper contains no analysis of where Bitcoin’s electricity comes from. Yet the claim that Bitcoin is fossil-fuelled circulates widely, often attributed to him. Four papers cite him directly for the claim that Bitcoin’s energy use will grow without bound — a projection his model specifically refuted.
The most consequential misattribution concerns climate harm. De Vries’ paper contains no estimate of Bitcoin’s carbon emissions. Yet 185 papers assert that Bitcoin is a significant driver of climate harm, and 50% of them do so via citation chains tracing back to de Vries through intermediary papers. His name is attached to a conclusion he never reached.
§3 — The silence
There is a body of peer-reviewed literature that questions de Vries’ methodology. It has been available for years. We checked all 6,679 papers in our full dataset for any sign of engagement with it.
3 engaged substantively. Zero in 2025.
For policymakers, this matters directly. When a regulator cites “the academic literature” on Bitcoin’s carbon footprint, they are citing a literature that is 69% inherited figures, 0.1% engaged with methodological critique, and 92% stripped of the uncertainty language its original author included.
§4 — The numbers and where they came from
When papers report a specific figure for Bitcoin’s total energy consumption, 69% of the time that number comes from another paper — not from original measurement. We found 26 papers with independent, original estimates. The range is striking:
The citation chain is not just passing around an old number. It is systematically understating Bitcoin’s current energy consumption by more than an order of magnitude, in the papers most likely to be read by people making policy.
§5 — What this means
This analysis is not an argument that Bitcoin’s energy use is acceptable, or that environmental concerns are unfounded. The independent measurements show real consumption at substantial scale. The concerns are legitimate.
What this analysis shows is that the evidence base underlying specific policy claims — that Bitcoin is predominantly fossil-fuelled, that its energy demand is unbounded, that it is a driver of climate change — has not been independently verified at the scale the policy stakes require. These claims have hardened into fact through repetition, not through evidence.
The appropriate response is not to dismiss the concerns. It is to demand better evidence before acting on the ones that have not been established.
This analysis is part of DARI’s ongoing research programme on the epistemic quality of cryptocurrency-related academic literature. The peer-reviewed version — “Proposition Inheritance in the Bitcoin Energy Literature” — is under submission. Research pre-registered at OSF prior to production analysis. Classification performed using a validated LLM-based systematic review methodology (MiniMax M2.7 + Claude Haiku 4.5). Inter-rater reliability (Cohen’s κ) computed against a 28-paper human-coded anchor set.