freed 2018-05-24, from a root note on the economics of rented hashpower · grew beside Rindex: the Robustness Index

The Realistic, Lucrative Case of an Ethereum Classic Attack

In May 2018 I asked a simple question: what does it cost to rewrite Ethereum Classic, and what does it pay? The market treated 51% attacks as folklore — a thing that happens to toy chains, not to a top-20 asset. The numbers said otherwise. The numbers said the attack was a trade, and the trade was open.

The argument had three legs:

1. Security is a budget, not a property. A proof-of-work chain is exactly as safe as the honest hashrate you must out-buy. ETC ran on a small fraction of Ethereum's hashrate while carrying exchange-grade liquidity.
2. Hashpower became a spot market. Rental markets like NiceHash removed the capital barrier. An attacker no longer buys machines; he rents an hour of majority.
3. Exchanges price risk in minutes. Confirmation windows were tuned for the cost of attacking Ethereum, not Ethereum Classic. Deposit, trade out, rewrite the hour, keep both sides.

The conclusion wasn't that ETC might be attacked. It was that the attack had positive expected value, and things with positive expected value eventually get done.

The exact figures — hashrate ratios, rental cost per hour, the double-spend take that clears the rental bill — are in the Medium original. The shape of the argument is what this note preserves: when the cost of rewriting history drops below the value sitting inside a confirmation window, history gets rewritten.

Last tended 2019-01-10 — it happened

On 2019-01-05, Ethereum Classic was 51%-attacked. Coinbase identified deep chain reorganizations with double spends totaling roughly $1.1M. Exchanges froze ETC deposits and stretched confirmation times — the exact repricing the piece said was overdue. I take no pleasure in the validation. I take the lesson: read the budget, not the brand.

Press that covered it

• Investopediacited the cost model after the January 2019 attack2019
• Cointelegraphcovered the prediction and its confirmation2018–2019
• The Next Web“this attack was called eight months in advance”2019
• CryptoBriefinganalysis feature on rentable-hashrate risk2018