A 2010 bug briefly created 184 billion bitcoin

Inherited archive entry · retained from the earlier edition. Its factual, legal and policy claims have not all been reverified for 19 September 2026. Research scope.

What the block contained

On 15 August 2010, block 74638 included a transaction with two outputs of roughly 92.2 billion bitcoin each, far beyond the 21 million coin design cap. The forum thread 'Strange block 74638' shows the community diagnosing the figure the same afternoon: developer Jeff Garzik flagged the exact number as suspiciously close to a 64-bit integer limit, and another participant, posting as lfm, identified the cause within hours, writing that the code checked each transaction output for negative numbers individually but forgot to check that the sum of two outputs together could also be negative. Two large, carefully chosen output values summed past the range the validation code checked, wrapping around to a total that passed as valid.

How the fix worked

The patch committed that day, titled 'fix for block 74638 overflow output transaction', adds an explicit constant for the maximum possible supply and checks both individual output values and their running total against it before accepting a transaction. This is a narrow, specific correction: it does not change how mining or difficulty works, only what counts as a valid transaction. Within hours of the block appearing, a patched client was available, and nodes running it refused to extend the chain built on the flawed transaction, instead continuing to build on the last valid block before it, a reorganisation that every node choosing to run the patch had to accept.

What enforced the correction, and what that depended on

Nothing about proof-of-work automatically excludes an invalid transaction; the miner that included it had already done the work to produce a block. What excluded it was every other node's independent validation rule rejecting that block regardless of the work behind it, exactly the distinction the Bitcoin Wiki's account (cited here only as a secondary summary, for the total figure and the specific reorganisation height) describes as reverting the chain past block 74691. The correction required operators to notice, agree a fix was correct, and actually run it, a coordination exercise, not an automatic property of the protocol.

• What other consensus rules exist only because every node happens to run compatible software?
• How quickly would a comparable bug be noticed and patched today, with far more value at stake?
• Does a cap enforced by code remain absolute if enough participants choose not to enforce it?

The supply cap held in August 2010 because people fixed a specific bug quickly and in agreement, not because the cap is immune to bugs; that distinction is worth keeping in view whenever the 21 million figure is described as self-enforcing.