Stellar's homepage advertises, in a block headed "Twelve years in production", a settlement time of "5s". Over the 55,000 consecutive ledgers ending at 06:18 UTC today, the network closed every one of them exactly 5.000 seconds apart. Not a mean of five seconds. All 54,999 intervals, identical.
That claim was not true in September. Bisecting the chain gives ledger 64,458,445, closed at 17:00:01 UTC on 16 September, as the last under protocol 27, and ledger 64,458,446 five seconds later as the first under protocol 28. Before it, close times ran five, six, seven and occasionally eight seconds; the 2,000 ledgers immediately preceding averaged 5.60, and the 412-day mean is 5.65. The round number became literally correct five weeks ago, and has held through the protocol 29 upgrade the chain now runs.
Consensus agrees these timestamps, and Stellar's documentation warns they may lag or run ahead of real time, so we checked against wall clock: 102 ledgers in 510 real seconds, with 0.2 seconds of skew across the test. A second independently operated indexer returned identical counts across 2,000 ledgers.
What the five seconds carries
The same window settled 11,509,120 successful transactions in 274,995 seconds, which is 41.85 a second, and 27,757,262 operations, 100.94 a second. Stellar markets operations, a ratio of 2.41 to one, so the unit alone moves the headline two and a half times. A narrower 5,000-ledger cut gives 45.38 transactions a second, eight per cent flattering, which is why the wide window is quoted.
It also turned a great deal away. Another 7,198,978 submissions failed, 38.48 per cent of everything offered. The ledger header shows why: there is a ceiling of 1,000 operations per ledger, and 11.35 per cent of ledgers submitted sets at or over it. The chain runs at roughly half that ceiling on successful work while rejecting two in five.
A sample of 35,355 operations across the window runs roughly 34 per cent payments, 37 per cent order-book management and 17 per cent contract calls, give or take a couple of points depending where you cut it. Around 92 per cent of the order operations amend or cancel one that already existed, and a single account is 11.3 per cent of the sample, its sell and buy offers within a few per cent of each other — a market maker requoting rather than anybody trading.
What the money says
DefiLlama puts total value locked at $229.0m, eighteenth of 421 chains. Every decentralised exchange on Stellar traded $541,133 in 24 hours, 98.7 per cent of it on Aquarius alone and down 54 per cent on the week. Fees across all protocols came to $79,601, of which the chain kept $1,000; measured from fee-pool deltas in the headers, $886 a day. XLM trades at 12.60p, a capitalisation of £4.44bn.
The supply figures disagree. The ledger header reports 105,443,902,087 XLM in existence against CoinGecko's 50,001,786,837 total, a gap almost exactly the stock the foundation reported burning in 2019 and which the header's counter never deducted. The quoted capitalisation prices only the 35.2bn CoinGecko counts as circulating; price the header's number and you get £13.3bn, three times as much.
Five seconds, held to the millisecond, across 55,000 consecutive ledgers. And 38 per cent of what was offered did not get in.
Who can stop it
Stellar has no staking and no slashing, so decentralisation here is a question of whose votes count. The default quorum configuration published by the foundation's own validator names ten organisations, each contributing three validators, with a threshold of seven: seven must agree to close a ledger and four can block one, so thirty validators out of 103 carry the decision. Naming them is more than most chains manage, but it does not make four-of-ten halt power less concentrated, and at least two of the ten also issue assets on the chain they secure.
Bear and bull
The bear case is a payments network whose payments are thin. A chain earning under $1,000 a day in fees and holding $229.0m of locked value is not short of block space. It is short of users.
The bull case is that the engineering did what it said. A deterministic five-second close is useful for settlement, it arrived as a shipped upgrade rather than a promise, the transaction cost we measured is $0.000245, and the quorum members are named and regulated. Whether regulated finance routes volume through it, ledger headers cannot say.
| Advertised settlement time | "5s" |
| Measured, 55,000 ledgers | 5.000 s, every interval |
| Measurement window | ledger 65,086,947 to 65,141,946, 3.18 days |
| Mean before protocol 28 | 5.60 s (412-day mean 5.65 s) |
| Transactions per second | 41.85 successful (100.94 operations) |
| Failed share of submissions | 38.48% |
| Ledgers at or over the 1,000-op ceiling | 11.35% |
| Total value locked | $229.0m |
| XLM price, market capitalisation | 12.60p, £4.44bn |
Close times, throughput, operation mix, account concentration, fee-pool deltas and supply measured from Stellar mainnet via horizon.stellar.org and cross-checked against horizon.stellar.lobstr.co, window ledger 65,086,947 to 65,141,946. Claims from stellar.org and developers.stellar.org; value locked, exchange volume and fees from DefiLlama; price from CoinGecko; quorum set from the foundation validator via the OBSRVR Radar API, node records last refreshed 14 October 2026. All retrieved 26 October 2026, 06:18–06:30 UTC. Contested: the ledger header's 105,443,902,087 XLM against CoinGecko's 50,001,786,837, both above, and the 2019 burn's effect on the header is unverified. Shares are from a 35,355-operation sample, not all 27.76m, and are rounded accordingly: a contiguous seven-minute sample gave 12 per cent payments against 34, so composition varies sharply with when you look. The 11.35 per cent ceiling figure is measured on submitted transaction sets; on applied operations it is 0.63 per cent. Not a recommendation to buy, sell or hold anything.