This file opened on 3 January 2009, with the first block of the chain, because coins have been going missing for exactly as long as coins have existed. The question is how many. The most cited answer comes from Chainalysis, which estimates that between 2.3 and 3.7 million bitcoin are likely lost for good — the range priced live at the top of this page.
The Bureau insists on showing its work, because this number is an estimate and is often repeated as if it were a body count. The method: count coins that have not moved in five or more years, sitting in wallets from the early era that have never spent a single output, and treat them as probably gone. That's a dormancy proxy, not a confirmed death. Some of those coins belong to the patient. About 1.1 million of them belong to Satoshi Nakamoto, untouched since 2010 — the largest unclaimed certificate on earth, and nobody knows which side of the ledger to file it on.
Run the proxy anyway and you get the famous headline: roughly twenty percent of all bitcoin that will ever exist may already be unreachable. One coin in five, give or take a Satoshi.
Where do lost coins go? Nowhere. That's the strange part. They stay on the chain, fully visible, forever — you can look at them, point a block explorer at them, watch the world reprice them hourly. They are not destroyed; they are stranded. And every stranded coin makes the remaining ones scarcer, which means the people who lost their keys have been, collectively and involuntarily, the most generous benefactors the holders have ever had.
The evidence for the estimate is filed throughout this Archive: 8,000 under a landfill in Newport (Case 002). 7,002 behind two remaining guesses in a Swiss vault (Case 003). A share of Mt. Gox's 850,000 that no trustee ever found (Case 006). The estimate is abstract. The files are not.
The Bureau's finding: UNRESOLVED by definition — the missing fifth gets an estimate, never a verdict. Some losses get a number, though. Yours, for instance. The calculator is below, and unlike this file, it closes.