Ordinals and Inscriptions: Bitcoin NFTs and Their Effects on the Network

Ordinals are a numbering scheme that gives every single satoshi a number. Inscriptions are data, such as an image or a text, written onto such a numbered satoshi and stored permanently in the blockchain. Together they make Bitcoin NFTs. Since Casey Rodarmor launched the project in January 2023, images and tokens have competed with ordinary payments for block space. At times this drove up fees, brought miners income and set off the fiercest dispute over Bitcoin software in years, which as of November 2026 has not been settled.

Ordinals: every satoshi gets a number

One bitcoin consists of 100’000’000 satoshis. To the Bitcoin network all satoshis are the same, just as two one-euro coins are worth the same. The blockchain only knows amounts in outputs, not individual satoshis.

Casey Rodarmor, a former Bitcoin Core contributor, proposed ordinal theory in 2022: every satoshi gets a sequential number in the order it was mined. The first satoshi from the first block is number 0. When an output is spent, the satoshis move along by a fixed rule: the first satoshis of the inputs go into the first outputs, in order ("first in, first out").

That way every satoshi can be traced through the entire history. The result is a kind of collectors' market. Rodarmor defined rarity tiers, for example:

Important: these numbers are written nowhere in the blockchain. They arise when software like the ord program recomputes the entire history. Anyone who doesn't apply the convention sees only ordinary bitcoin. How satoshis work as a unit is explained in What are sats?.

Inscriptions: writing data onto a satoshi

A number alone isn't an NFT yet. That takes content. Inscriptions write arbitrary data into a transaction: an image, a text, HTML, audio. The data is then considered "inscribed" on the first satoshi of the output. Whoever owns that satoshi owns the inscription.

Technically this takes two transactions:

1. Commit: A payment to a Taproot address whose hidden script contains the data.
2. Reveal: This output is spent via the script path. The script has to be revealed, and with it the data.

The data sits in a section that is never executed: OP_FALSE OP_IF … OP_ENDIF. To the network this is a valid, if odd, script. The technical background, tapscript and script paths, is explained in Taproot and Schnorr. What a Bitcoin script can do at all is covered in Bitcoin scripts and smart contracts.

The first inscription, number 0, is a pixel skull image inscribed in December 2022. On 21.01.2023 Rodarmor released the software for mainnet, and within a few weeks tens of thousands of inscriptions appeared. By now there are several tens of millions.

The witness discount: why images on Bitcoin became cheap

Inscriptions grew so fast because of a rule from the 2017 SegWit upgrade. Since then a transaction's size has been measured in weight units. Normal data counts 4 units per byte, data in the witness (signatures and scripts) only 1 unit. A block may have 4’000’000 weight units.

The discount was meant for signatures. It was supposed to price transactions with many signatures more fairly and make cleaning up many small outputs cheaper. Inscriptions put their data exactly there and so pay only a quarter.

Taproot in 2021 also removed the old limit of 10’000 bytes per script. An inscription can therefore fill up to the limit of a whole block. On 01.02.2023 the pool Luxor found block 774’628 at 3.96 MB, almost entirely filled with a single image from the "Taproot Wizards" collection.

BRC-20: tokens on inscriptions

In March 2023 an anonymous developer using the pseudonym "domo" invented BRC-20. The idea: tokens whose rules are written in small text inscriptions in JSON format. A "deploy" inscription creates a token, "mint" issues an amount, "transfer" prepares a transfer.

The Bitcoin network checks none of this. Who owns how many BRC-20 tokens is computed by indexers, programs outside Bitcoin that evaluate all inscriptions according to the rules. If two indexers disagree, there is no referee.

The first token, "ordi", set off a wave. Anyone who wanted to mint had to be quick and paid accordingly. In May 2023 the median fee rate rose to 436 sat/vB, and the exchange Binance temporarily paused withdrawals. The details of that spike are in Mempool analysis. A successor that binds tokens directly to outputs rather than indexer accounts arrived in 2024 with the Runes protocol.

Effects on the network: fees, block space, nodes

Fees: Inscriptions bid for the same block space as payments. In several waves in 2023 and 2024 fees therefore rose sharply. For savers with many small amounts, consolidating became expensive. Since 2025 interest has faded markedly and fees were low for a long time. What drives fees in general is explained in Bitcoin transaction fees.

Miner revenue: In the waves, fees made up a large part of the reward. Those worried about the security budget after the halvings see in it a foretaste of how demand for block space can pay miners. That debate is covered in What comes after the last halving?.

Node storage: A node stores every block, so every image too. Full blocks make the blockchain grow faster. Nodes can discard witness data after validation ("pruning"), but archive nodes keep it.

Many small outputs: Every inscription sits on its own small output, often 546 or 10’000 satoshis. BRC-20 transfers also create outputs that nobody will ever economically spend. The set of all unspent outputs (the UTXO set), which every node keeps in memory or on a fast disk, has grown significantly as a result.

The dispute: Bitcoin Core, Bitcoin Knots and OP_RETURN

Resistance to inscriptions formed early. Critics such as developer Luke Dashjr, maintainer of the node software Bitcoin Knots, see them as abuse of a loophole: the witness discount was meant for signatures, not images. At the end of 2023 a vulnerability identifier was assigned for it (CVE-2023-50428): data limits could be bypassed by disguising data as never-executed code.

What Knots does: Bitcoin Knots filters inscriptions and other data transactions out of its mempool. A Knots node doesn't relay them. It still accepts them in blocks, because under the consensus rules they are valid.

What Core does: Bitcoin Core's developers consider such filters ineffective. Their arguments: anyone who wants to store data finds a way, if necessary directly through miners, bypassing the public mempools. Worse are methods that pack data into fake addresses; nobody can ever spend such outputs, and they stay in the UTXO set forever. With Bitcoin Core 30 in October 2025, Core raised the default limit for OP_RETURN outputs from 83 bytes to around 100’000 bytes. OP_RETURN data is marked unspendable and doesn't burden the UTXO set. But it pays the full price of 4 weight units per byte.

The reaction: Many node operators switched to Knots in protest. According to Coin.dance's count, more than a fifth of publicly reachable nodes ran Knots in July 2026. Other counts arrive at considerably less, and the figures are regarded as uncertain.

The soft fork attempt: In 2026 the proposal BIP 110 was meant to limit data in transactions for one year by consensus rule. It shipped in Knots, but almost no miner signalled support. On 08.08.2026 the BIP 110 nodes split off onto a chain of their own that stopped after two blocks. The sequence of events is described in Bitcoin soft fork explained.

Both programs follow the same consensus rules. A block with inscriptions is valid for both. The dispute is about what a node should relay, and about what Bitcoin is for. How nodes and mempools handle such rules is explained in The mempool.

What ordinals mean for you as a saver

Fees: A new craze can make block space expensive again. Consolidate small amounts in quiet periods, not during a hype.

Wallets: If you receive satoshis carrying an inscription, a normal wallet doesn't know. It may spend that output as part of a fee, and the inscription goes to the miner. If you don't care, no problem. If you want to keep it, you need a wallet that understands ordinals, or coin control to freeze individual outputs.

Value: Most collections and BRC-20 tokens have lost a large part of their price since their peaks. An NFT on Bitcoin isn't bitcoin. Its price depends entirely on collectors.

Network: Nothing changes for the security of your coins. The debate concerns relay rules and storage, not the 21 million or your keys.