What Are Bitcoin Ordinals?
A numbering scheme that lets data be pinned to a single satoshi. Nearly 128 million inscriptions had been made by October 2026, and the block space they use is still argued over.
TL;DR
- Ordinal theory numbers every satoshi in the order it was mined, so a single sat can be followed from one transaction to the next.
- An inscription attaches data such as an image or text to one sat, stored in a transaction's witness, which SegWit discounted in 2017 and Taproot made roomy in 2021.
- Casey Rodarmor's ord software opened mainnet inscriptions in January 2023, and BRC-20 tokens followed in March, helping push some blocks' fees above the subsidy that May.
- Supporters see paying users and fee income for miners, while critics see non-payment data crowding out ordinary transfers and adding to every node's storage.
On 1 February 2023 the mining pool Luxor produced block 774,628, and at 3.96 MB it was the largest block Bitcoin had carried up to then. It held 63 transactions, one of which took up about 99% of the space, and nearly all of that was a single image of a wizard, the first piece of a collection called Taproot Wizards.
No consensus rule was broken. The block was valid, and every node on the network accepted it. But it hardened an argument among Bitcoin users about what the chain is for, and that argument is still running more than three years later.
The wizard was an ordinal inscription: a piece of data tied to one specific satoshi. This is an educational guide, not financial advice.
Every satoshi gets a serial number
A bitcoin divides into 100 million satoshis, or sats, and the network's own rules never tell one sat apart from another. Ordinal theory does. It numbers each sat in the order it was mined, then follows it through every transaction on a first-in, first-out basis, so any individual sat can be traced from the block that created it to whichever wallet holds it today.
None of this needed a change to Bitcoin. The numbering is a convention, a way of reading the existing chain, and it is applied by ord, the software written by Casey Rodarmor. He made the first inscription himself on 14 December 2022, in block 767,430. On 20 January 2023 he released ord 0.4.0 and told users it could now make inscriptions on Bitcoin's main network that would not break in a later protocol change.
It took less than a fortnight for someone to fill a block, and by early October 2026 the explorer at ordinals.com counted nearly 128 million inscriptions.
Where the data actually sits
An inscription is made in two steps: a first transaction, the commit, locks some bitcoin to a Taproot output that hides a script. A second, the reveal, spends that output and publishes the script, with the content wrapped inside it in an "envelope", a stretch of code built so that it never runs. Bitcoin checks the spend, ignores the content and stores the lot.
That content lands in the witness, the part of a transaction that normally holds signatures. Two upgrades made the witness a cheap and roomy place for it.
SegWit, activated on 24 August 2017 at block 481,824, introduced block weight. Ordinary transaction bytes count four times towards a limit of 4 million weight units, while witness bytes count once. A block packed with witness data can therefore approach 4 MB, against about 1 MB for one with none.
Taproot, which went live on 14 November 2021 at block 709,632, dropped the old 10,000-byte cap on scripts for its new script type, leaving the block weight limit as the only consensus ceiling.
Put the two together and one image can fill nearly an entire block, which is exactly what happened in block 774,628. Standard relay rules cap a single transaction at about a tenth of a block, so the wizard went straight to the miner, Luxor, and was included with no fee attached. Inscriptions that take the normal route pay the same discounted rate per byte as signatures do.
BRC-20 and the fee spike of May 2023
On 8 March 2023 a pseudonymous developer called domo published BRC-20, which used text inscriptions to create tokens. Each one is a short snippet of JSON (a plain-text data format) that deploys a token, mints some of it or transfers it. The first token, ordi, was deployed the same day, and domo called the whole thing an experiment that should not be taken as the standard for tokens on Bitcoin.
Then came the minting rush, and by 8 May BRC-20 activity made up more than half of all transactions on the network, roughly 488,000 transactions were waiting unconfirmed, and Binance had paused bitcoin withdrawals twice within a day, blaming congestion. In the week to 10 May, fees paid to miners came to 33% of the new-coin subsidy, against 6% the week before, according to Luxor's Hashrate Index, and some blocks paid out more in fees than the 6.25 BTC subsidy itself.
Ordinary payments paid those rates too, because miners fill blocks with the transactions offering the highest fee per unit of weight, so a transfer and a token mint bid for exactly the same room. A later token system, Runes, arrived with the April 2024 halving block.
The argument over block space
Supporters start from the rules: inscribers usually pay the going fee like anyone else, and Bitcoin's consensus rules treat a payment and a picture alike. Miners need the income too, because the new-coin subsidy shrinks at every halving and the original design has fees eventually taking over, so to them inscriptions are paying customers.
Critics start from the purpose: they see Bitcoin as a payment network whose blocks are scarce by design, argue that the witness discount was meant for signatures rather than images, and point out that every full node has to download and check the extra data. When inscription demand spikes, they say, ordinary users pay more to move their own money.
The dispute even reached the vulnerability database: on 9 December 2023 the US National Vulnerability Database published CVE-2023-50428, which describes inscriptions as a way around node software's size limits on embedded data and notes that what the Bitcoin Knots project calls a vulnerability, others consider "not a bug". The entry is tagged as disputed.
The argument spilled into node software as well: Bitcoin Core 30.0, released on 10 October 2025, raised its default cap on data in OP_RETURN outputs, a separate slot for arbitrary data, from 83 bytes to 100,000 bytes. The fight then reached the rules themselves, with BIP-110, a soft-fork attempt to limit non-payment data, failing when its nodes split off in August 2026 after only 2.53% of blocks had signalled support.
What it means if you only send bitcoin
Mostly, fees. A payment and an inscription bid for the same block space, so when minting surges, as it did in May 2023, the price of a place in the next block goes up for everyone. Send with too low a fee during a rush and the payment can sit unconfirmed until the queue clears.
Holders of inscriptions face a stranger risk: the inscription lives on one particular sat inside one particular unspent output, and a wallet that was never built to recognise ordinals sees only bitcoin there. Spend from that wallet and it may grab the output, sending the inscription to someone else or spending it as part of the fee, and the ord documentation warns about exactly this.
People often call inscriptions Bitcoin NFTs, and the label mostly fits. One difference the ord documentation stresses is where the content lives: entirely on-chain, inside Bitcoin's own blocks.
Frequently Asked Questions
Ordinals are serial numbers given to individual satoshis, the hundred-millionth units of a bitcoin, under a scheme called ordinal theory. Sats are numbered in the order they were mined and tracked first-in, first-out through transactions. People also use the word loosely for inscriptions, the images, text and other data attached to particular numbered sats.
They behave much like NFTs. An inscription ties a piece of content to one specific satoshi, and whoever controls that sat controls the inscription. The content is written into a Bitcoin transaction rather than kept elsewhere, and no separate token, sidechain or change to Bitcoin's rules is needed for it to work.
Blocks have a fixed weight limit and miners choose the transactions paying the highest fee rate. When inscription and BRC-20 minting surged in May 2023, about 488,000 transactions were queued unconfirmed on 8 May, and some blocks paid miners more in fees than the 6.25 BTC subsidy. Ordinary payments had to outbid that demand.
An ordinal inscription attaches a unique piece of content to one satoshi, so each is distinct. BRC-20, published on 8 March 2023 by a pseudonymous developer called domo, uses text inscriptions holding short JSON instructions to deploy, mint and transfer interchangeable tokens. Its first token was ordi, and its creator described the format as an experiment.
Yes. An inscription sits on a specific sat inside a specific unspent output, and a wallet that cannot recognise ordinals treats it as ordinary bitcoin. If that wallet picks the output when you send funds, the inscription can go to the recipient or be spent as fees. The ord documentation advises against sending bitcoin from such a wallet unless you choose the coins manually.
About the author — Dan Clarke
Dan Clarke is the author of Bitcoin: The Complete Guide and a former content lead at Binance Academy, where he wrote crypto education for readers arriving with no background in the subject. He has worked in the cryptocurrency industry since 2017. His rule for these guides: plain language first, precision where it matters, no cheerleading.