100% uptime since block 0.
The Bitcoin network has never gone down. Since block 1 was mined on 9 January 2009 it has never been switched off, paused for maintenance, or closed for a weekend or a holiday. That is what 100% means here. It does not mean nothing has ever gone wrong: twice, the chain itself had to be corrected, and the network kept running through both.
Since block 1, mined on 9 January 2009.
No scheduled maintenance. Ever.
August 2010 and March 2013. The network kept running through both.
Why it never goes down
Bitcoin runs on tens of thousands of independent nodes, each holding a full copy of the ledger, each independently validating every transaction. There is no central server. There is no authoritative data centre. If half the nodes went offline tomorrow, the network would continue operating from the other half. If 99% went offline, the remaining 1% would still process blocks and maintain consensus.
This is the structural reason for the uptime. It is not that Bitcoin has exceptionally good engineering (though it does). It is that there is no single thing to take down.
What 100% does not mean
It does not mean a block every ten minutes without fail. Blocks arrive about every ten minutes on average, and in the early months, when a handful of people mined on home computers, the gaps could be far longer. The chain’s own timestamps show two stretches in 2009 when more than a day passed without a block: 22–23 May and 5–6 June. Nobody switched the network off, and anyone could have mined a block; there was simply too little mining to guarantee one.
Nor does it mean nothing has ever gone wrong. Twice in its first five years a software bug forced the network to abandon the chain it was on and converge on a corrected one. Blocks were mined throughout both, but some of them were thrown away.
15 August 2010: the overflow bug
Block 74,638 contained a transaction that created 184,467,440,737 bitcoins out of nothing, paid as two outputs of 92.2 billion each. The code that checked a transaction’s outputs did not allow for values so large that their sum overflowed. It was spotted the same day. Within five hours of the discovery a fixed version of the software was out that rejected the transaction, and miners who upgraded started a corrected chain from the block before it. At block 74,691, on the morning of 16 August, the corrected chain overtook the bad one and every node followed it.
The 184 billion coins, and the transaction that made them, do not exist in the chain today. The episode is still visible in it: block 74,637 is timestamped 17:02 UTC on 15 August, and the next block, 74,638, 23:53 UTC. The hours before the fix, when the whole network was building on the bad chain, are simply missing.
11–12 March 2013: the database limit
A miner running the new version 0.8 produced block 225,430: valid, but with more transaction inputs than earlier versions’ database was configured to handle. Nodes running 0.8 accepted it; older nodes rejected it, and the chain split in two, with around 60% of mining power on the 0.8 side. To restore a single chain, two large mining pools running 0.8, BTC Guild and Slush’s pool, switched back to the older version at significant cost to themselves. Twenty-four blocks later, at block 225,454 early on 12 March, the chain without the large block pulled ahead, and the 0.8 nodes reorganised onto it.
During the split, at least one large payment was double-spent, by someone experimenting to see whether it could be done rather than trying to steal, according to the developers’ post-mortem.
What the two have in common
Both were bugs in the software, not attacks on the network. Both were caught within hours, and both were resolved the way anything on the network changes: enough of the people running it chose to run corrected software. Neither is a promise that it cannot happen again, and neither took the network down.
Compared to other networks
Visa and Mastercard have had multiple multi-hour outages over the last decade. CHAPS closes at weekends and on bank holidays, and the Fedwire system in the United States is available approximately 22 hours per day, five days a week. Even SWIFT, whose messaging network is built to run around the clock, schedules maintenance windows, and a payment sent through it on a Friday night usually waits for the banks at either end to open on Monday. Individual banks experience regular scheduled downtime for maintenance.
These are not failures. They are architectural choices. A centralised system must be taken offline periodically to be maintained. A decentralised system has no single point to maintain.
What it means for you
When you hold bitcoin in self-custody, you can transact at any time. Sunday at 3am on Christmas Day. Across a bank holiday. During a banking crisis. During a government shutdown. The network does not care what day it is, what jurisdiction you are in, or what the political situation looks like. It keeps adding a block roughly every ten minutes, whatever the day.
For most people, this is a property they will never need to rely on. For people in countries with capital controls, unstable banking systems, or regulatory uncertainty, it is the most important property of all.
Sources
- Value overflow incident, 15 August 2010 (Bitcoin Wiki)
- CVE-2010-5139
- BIP 50: March 2013 chain fork post-mortem
- 11/12 March 2013 chain fork alert (bitcoin.org)
- Block heights and timestamps: the chain itself, via the linked blocks on mempool.space.