Bitcoin node counters go dark following Bitnodes expiry

It used to be easy to estimate how many Bitcoin full nodes were easily reachable over the internet, but almost all of those counters have gone offline this month.

As of Tuesday, Coin Dances node counter reported a last-updated time of 476 hours ago. Indeed, the page displays 23,795 reachable nodes as of its latest update last month.

Bitcoin node estimates plastered across dashboards, glossaries, and other webpages havent moved in roughly three weeks.

The main culprit seems to be Bitnodes.io, a popular crawler that ran continuously since 2013 yet went dark after its domain expired on May 3, 2026. A single developer, Addy Yeow, ran the project with early financial support from Balaji Srinavasans 21 Inc, once the best-funded company in the Bitcoin industry.

Neither Srinavasan nor Yeow have issued public statements on the GitHub repository or anywhere else about the outage.

NewHedge used Bitnodes API and, therefore, has taken its counter offline. Another Bitnodes API user, SatoshiDashboard, also went down.

Clark Moody‘s dashboard simply re-posts Bitnodes’ old cached number.

Navigating to bitnodes.io itself returns a DNS error.

Bitnodes was a fan favorite

Two weeks into the outage, BitMEX Research proposed a rudimentary replacement crawler, Bitnod.es, using off-the-shelf tooling.

With sincere modesty, BitMEX Research conceded that the substitute is significantly inferior to the original.

Indeed, Bitnodes ran continuously for 13 years. The well-capitalized 21 Inc sponsored it at first, including under 21 Incs 2017 rebrand to Earn.com.

The timing of the outage is particularly unfortunate given widespread uncertainty about how many nodes are operating the Bitcoin network currently.

How many Bitcoin nodes are online, 14,000 or 350,000？

IP and onion addresses of nodes have become particularly hard to crawl over the last month.

For example, Jameson Lopp recently flagged a roughly four-fold spike in IP-like addresses propagating across Bitcoins ADDR (address) gossip messages.

Prior to this month, a research group of the Karlsruher Institut für Technologie in Germany found roughly 60,000 daily unique IPs operating the Bitcoin network. By early May, they detected a spike to roughly 250,000.

Their same dashboard has continued to skyrocket with an estmate now near 350,000.

Amid outages at mainstream counters, users are finding alternatives with wildly different numbers than Bitnodes.

Luke Dashjr‘s tracker, for example, shows 72,682 Bitcoin Core nodes and 18,724 Bitcoin Knots nodes — far higher than Bitnodes’ combined estimate in the 20,000s in April.

Another indie tracker actually estimates a sharp decline in Bitcoin listening nodes in April from roughly 18,000 in March to less than 15,000 in April and under 14,000 this month. Bitnod.es finds less than 2,500.

Estimating Bitcoin nodes by oneself

Bitcoin‘s node-counting infrastructure was always thin, but it’s now entirely stale and poorly maintained.

As billionaire Srinivasan became preoccupied with acquiring an island near Singapore instead of supporting Bitnodes, and indie developers ran short on funds for operating public reporting of Bitcoin infrastructure, the only way to find out how many nodes are online has become DIY.

Anyone can run a Bitcoin node, start connecting to peers, and gossip with other operators to guess about additional IP, onion, or IP-like addresses that might be online.

Although connecting over the mesh network to conduct an estimate takes time and will yield a different absolute number than anyone else, due to the voluntary and random nature of global participation, its possible to arrive at a reasonable estimate with sustained effort.

Unfortunately, theres no financial incentive other than reporting for the public good to extensively and diligently connect to as many peers as possible.

Most node operators are entirely content to have a couple hundred or thousand connections, which are sufficient for validating blocks and forthe purposes of receiving and broadcasting the majority of mempool-queued transactions.