Bitcoin Ordinals are a system that assigns a unique number to every satoshi on the Bitcoin blockchain, making individual sats identifiable, and then allows data such as an image, a text file, or a small program to be attached to a specific sat. That attached data is called an inscription, and the result behaves like a non-fungible token that lives entirely inside Bitcoin rather than on an external server.
The protocol has been running since December 2022 and has produced well over 90 million inscriptions. It has also caused the loudest argument in Bitcoin's recent history, an argument that escalated into an actual chain split in August 2026.
This guide covers what Ordinals are, how an inscription is physically created, how rare sats work, the one mistake that can permanently destroy an Ordinal, and where the market and the block space debate stand right now.
Key Takeaways
- Ordinals are numbers. Inscriptions are content. Ordinal theory numbers each satoshi in the order it was mined. An inscription is data attached to one of those numbered sats. People use "Ordinals" loosely to mean both, but the distinction matters.
- The data is stored on Bitcoin itself. Unlike most NFTs, which point to a file hosted elsewhere, an inscription's content sits inside the Bitcoin blockchain's witness data. Nothing can break the link.
- No protocol change was needed. Ordinals work using SegWit and Taproot, two upgrades built for entirely different reasons.
- You can destroy an Ordinal by accident. A standard Bitcoin wallet can spend an inscribed satoshi as an ordinary payment or as a mining fee. This is the single biggest practical risk and it is why dedicated Ordinals wallets exist.
- The market cooled, but the chain did not. Trading volume is a fraction of its 2024 peak, yet inscription counts kept climbing through 2025 and 2026.
- The fight over block space is real and unresolved. An attempt to restrict inscriptions at the consensus level, BIP-110, split off a minority chain on August 8, 2026 and stalled within hours.
What Bitcoin Ordinals Actually Are
Bitcoin Ordinals are a numbering scheme for satoshis combined with a method for attaching arbitrary content to individual numbered sats. A satoshi is the smallest unit of Bitcoin, worth 0.00000001 BTC, and there are 100 million of them in every bitcoin.
Ordinarily, sats are interchangeable. One sat is as good as any other, in the same way one cent is as good as any other cent. Ordinal theory removes that interchangeability by giving each sat a serial number based on when it was mined. Once a sat has an identity, it can be tracked, and once it can be tracked, content can be tied to it.
Compare that to how a typical NFT on another chain works. There, a smart contract holds a token ID and a pointer, usually a URL, to a file stored on IPFS or a company's servers. If that server disappears, the token still exists but the picture does not. A Bitcoin inscription has no pointer. The image bytes are inside a Bitcoin transaction, replicated across every full node that keeps the block. Casey Rodarmor, who built the protocol, calls these digital artifacts rather than NFTs, and the distinction is deliberate.
Three terms get used interchangeably and should not be:
- Ordinal: the number assigned to a satoshi.
- Inscription: the content attached to that satoshi.
- Digital artifact: the finished result, an inscribed sat that is complete, immutable, and owned outright.
What Is Ordinal Theory?
Ordinal theory is the set of rules that decides which satoshi is which. Rodarmor published it as a specification and later as the Ordinal Theory Handbook, which remains the reference document.
The rule is simple. Satoshis are numbered in the order they are mined, starting from zero. The first sat ever created is sat 0, the next is sat 1, and so on. When sats move between transactions, ordinal theory tracks them first in, first out, so the protocol always knows which sat ended up where.
The handbook defines several ways to write an ordinal number, and each reveals something different:
| Notation | Example | What it shows |
|---|---|---|
| Integer | 2099994106992659 | The raw sequence number |
| Decimal | 3891094.16797 | Block height, then position of the sat within that block |
| Degree | 1°1'0"0''' | Cycle, position in halving epoch, position in difficulty period, position in block |
| Percentile | 99.99971949060254% | How far through Bitcoin's total supply the sat sits |
| Name | satoshi | The number encoded using the letters a to z |
Degree notation matters most, because it makes a satoshi's rarity readable at a glance. More on that below.
Importantly, none of this required changing Bitcoin. Ordinal theory is a way of reading data that is already on the blockchain. Bitcoin nodes have no idea that Ordinals exist. Indexing software such as ord reads the chain and applies the rules, and the reference explorer at ordinals.com presents the result.
How Do Bitcoin Ordinals Work?
Inscribing is a two-transaction process that uses a feature of Taproot. Here is what actually happens.
Step 1: Prepare the Data
The file being inscribed, whether a PNG, a snippet of text, or an HTML page, is converted to hexadecimal and paired with a MIME content type so software knows how to render it later.
Step 2: Wrap It in a Taproot Script
The hex data is embedded inside a Taproot script. This is possible because BIP 342, part of the Taproot upgrade activated on 14 November 2021 at block 709,632, removed the strict size limit that previously applied to script elements.
Step 3: Broadcast the Commit Transaction
The commit transaction creates a Taproot output whose spending conditions are defined by that script. Critically, it only publishes a hash of the script. The content is not yet visible on-chain.
Step 4: Broadcast the Reveal Transaction
The reveal transaction spends the commit output and, in doing so, must publish the full script to prove the spend is valid. The content is now permanently on the blockchain. It lands in the witness portion of the transaction, and the inscription is made on the first sat of the first output.
Step 5: Wait for Confirmation
Both transactions sit in the mempool until a miner includes them in a block. Once confirmed, the inscription is part of Bitcoin's history and cannot be edited or removed.
Two design details explain why this is economically viable at all. SegWit, activated in 2017, separated signature data from transaction data and gave witness data a discount when fees are calculated. Taproot then lifted the script size ceiling. Together they made it roughly four times cheaper per byte to store content in the witness than in the main transaction body. Rodarmor's own account of how the pieces came together is worth reading on his blog.
A note on the timeline, because it is widely misreported. Rodarmor made the first mainnet inscription on December 14, 2022, a 793-byte pixel art skull in block 767,430, numbered inscription 0. He created it by hand, because the ord wallet's inscribe command had not yet been enabled on mainnet. That came on January 9, 2023, and Rodarmor publicly announced the protocol on January 20, 2023, with the mainnet launch generally dated to January 21, 2023. December 2022 is when Ordinals began. January 2023 is when they became available to everyone.
Ordinals, Inscriptions, BRC-20, Runes, and Rare Sats: What's the Difference?
This cluster of terms causes more confusion than anything else in the topic. Here is each one in a sentence.
| Term | What it is | Fungible? | Depends on Ordinals? |
|---|---|---|---|
| Ordinal | The serial number given to a satoshi by ordinal theory | Not applicable | It is Ordinals |
| Inscription | Content attached to a numbered satoshi | No | Yes |
| Digital artifact | An inscribed sat, treated as a complete owned object | No | Yes |
| BRC-20 | Fungible tokens created by inscribing JSON instructions, introduced March 2023 by a pseudonymous developer known as Domo | Yes | Yes |
| Runes | A separate fungible token protocol by Rodarmor, launched at the April 2024 halving, using OP_RETURN instead of inscriptions | Yes | No |
| Rare sat | A satoshi considered special because of where it sits in Bitcoin's issuance history, inscribed or not | Not applicable | Uses ordinal theory to identify |
The practical takeaway: every BRC-20 token is an inscription, but almost no inscriptions are BRC-20 tokens. Runes are not inscriptions at all, despite constantly appearing in the same sentence. Rare sats can be valuable with nothing inscribed on them whatsoever.
One further concept worth knowing is recursive inscriptions, introduced in mid-2023. These let a new inscription reference data already on-chain instead of duplicating it. A generative art collection can inscribe a code library once, then have thousands of pieces call it, which cuts cost dramatically and makes complex on-chain artwork practical.
Rare Sats and the Rodarmor Rarity Index
Because ordinal theory knows exactly when every sat was mined, it can identify sats that coincide with notable moments in Bitcoin's issuance schedule. The Ordinal Theory Handbook defines six tiers, commonly called the Rodarmor Rarity Index.
| Tier | Definition | How many exist |
|---|---|---|
| Common | Any sat that is not the first sat of its block | Over 99% of all sats |
| Uncommon | The first sat of each block | One per block |
| Rare | The first sat of each difficulty adjustment period | One roughly every 2,016 blocks |
| Epic | The first sat of each halving epoch | One per halving, so five so far |
| Legendary | The first sat of each cycle, a cycle being six halvings | None yet issued |
| Mythic | The first sat of the genesis block | Exactly one |
Two details make this more interesting than a simple scarcity table.
First, the mythic sat can never be traded. The 50 BTC reward from the genesis block cannot be spent, due to a quirk in Bitcoin's original code, so the rarest satoshi in existence is permanently frozen where it is.
Second, no legendary sat exists yet. Cycles run roughly 24 years, so the first one is not expected until the 2030s.
Beyond the protocol's own definitions, collectors trade exotic sats, which are special by social consensus rather than by rule. Examples include sats mined in block 9, sats from the block Hal Finney mined, and sats from the 2010 pizza transaction. These categories are informal, which means their value rests entirely on whether other collectors agree they matter.
How Bitcoin Ordinals Compare to Ethereum NFTs
| Bitcoin Ordinals | Ethereum NFTs | |
|---|---|---|
| Where the content lives | Inside the blockchain, in witness data | Usually IPFS or a private server, referenced by a link |
| What creates it | Two standard Bitcoin transactions | A smart contract |
| Programmability | None at the base layer | Extensive, including royalties and upgradeable metadata |
| Creator royalties | Not enforceable on-chain | Enforceable by contract, though often bypassed in practice |
| Editions and collections | Handled by convention and indexers | Handled natively by the contract standard |
| Can the artwork break? | No | Yes, if the host goes offline |
| Consensus mechanism | Proof of work | Proof of stake |
| Typical cost driver | Bytes of data plus network fee rate | Contract complexity plus gas price |
Neither model is simply better. Ethereum's smart contracts give creators tools that Bitcoin cannot match, including automatic royalty payments and integration with DeFi protocols. Bitcoin gives up all of that in exchange for permanence and for settlement on the chain with the most accumulated proof of work.
The Biggest Risk: Spending Your Ordinal by Accident
An inscription lives on a specific satoshi, which sits inside a specific unspent transaction output. Your wallet sees that UTXO as ordinary bitcoin, because as far as Bitcoin's consensus rules are concerned, that is exactly what it is. When you send a payment, a standard wallet picks whichever UTXOs it likes to fund the transaction. If it picks the one holding your inscription, the Ordinal is either sent to whoever you paid or burned as a mining fee. There is no undo.
This is not a theoretical problem. It is why the Ordinals ecosystem built separate wallet software rather than reusing existing tools, and it is why the claim that Ordinals work smoothly with any Bitcoin wallet is wrong.
How to Handle an Ordinal Safely
- Use a wallet that understands inscriptions: Dedicated Ordinals wallets identify inscribed and rare sats and lock those UTXOs so they cannot be selected for ordinary spending.
- Keep inscriptions separate from spending money: Use a different address, ideally a different wallet, for sats you intend to spend.
- Check the UTXO before you send: A block explorer will show you which output holds what.
- Understand that marketplace trades use PSBTs: Partially Signed Bitcoin Transactions, defined in BIP 174, let a buyer and seller each sign one half of a trade so it settles atomically on-chain. You never hand custody to the marketplace. This is genuinely good design and it is worth knowing why your wallet asks you to sign what it asks you to sign.
- Watch for address poisoning and mempool sniping: Both are active attack patterns in Ordinals trading. Verify addresses character by character and be cautious with fee rates on high-value listings.
- Assume the marketplace is temporary: Custody of the asset is yours, but listings, indexes and interfaces are not. Keep your seed phrase, not just your account.
That last point stopped being hypothetical in 2026. Magic Eden, which at its height handled roughly 80% of all Ordinals and Runes trading and was the largest NFT marketplace in the world by monthly volume in March 2024, announced on February 27, 2026, that it was leaving Bitcoin to refocus on Solana. Its Bitcoin marketplace closed in the first week of March 2026, and its multi-chain wallet was discontinued on April 1, 2026. Users had to migrate assets on a deadline. Anyone who had treated the platform as storage rather than as a venue had a difficult few weeks. Trading has since redistributed to Bitcoin-native platforms, but the liquidity is thinner than it was.
Where the Ordinals Market Stands in 2026
The honest summary is that inscription activity and inscription trading have gone in opposite directions.
Inscriptions kept growing. The count has risen steadily throughout: roughly 52 million by the end of 2023, close to 70 million by September 2024, past 80 million in February 2025, and above 90 million by early 2026. More than 7.7 million inscriptions were created in the final quarter of 2025 alone, a period in which Bitcoin's price fell around 23%. People kept inscribing while the market fell.
Much of what gets called Ordinals activity is now Runes. Since the April 2024 halving, the Runes protocol has driven a large share of the metadata transactions and miner fees on Bitcoin, outpacing both Ordinals and BRC-20 by transaction share. Headline inscription counts include Runes and BRC-20 mints, so the raw number overstates how much art and collectible activity is actually happening.
Trading volume did not. Monthly Ordinals sales volume ran at approximately $53 million in January 2026, $33.6 million in February, and $46.8 million in March, according to CryptoSlam data. Set that against the peaks of early 2024 and the decline is severe.
The buyer base is small and real. March 2026 saw 59,585 Ordinals transactions involving 14,909 unique buyers and 11,768 sellers, at an average sale price of $785, with measured wash trading under 1%. That is a genuine collector market rather than a speculative bubble, and it is a fraction of the size it once was.
Fungible tokens shrank hard. The BRC-20 sector settled at roughly $71 million in market value with daily volumes near $23 million during the first half of 2026. ORDI, the first BRC-20 token, trades more than 95% below its all-time high.
Cumulative sales remain substantial. Lifetime NFT sales on Bitcoin approach $6 billion, which places the network third among all chains by that measure.
The most recent snapshot. In the week to August 29, 2026, global NFT sales across all chains fell roughly 45% to about $63 million, with Bitcoin NFTs accounting for the largest individual sales of the week, including one at $2.14 million. Bitcoin traded near $77,600 on that date. Ordinals are a small market now, but the high end still clears real money.
There is also a measurable footprint on Bitcoin itself. Taproot usage peaked at roughly 42% of transactions in 2024, driven overwhelmingly by inscription activity, and settled to around 20% by late 2025 as that activity slowed. The UTXO set passed 173 million entries in an April 2025 snapshot, more than triple its 2020 size, and close to 30% of those entries are linked to Ordinals inscriptions. Whatever you think of Ordinals, they changed the shape of Bitcoin's data.
The Block Space Argument: OP_RETURN, Core v30, and BIP-110
Ordinals have been contentious since week one, and in 2025 and 2026 the argument moved from opinion to code.
The disagreement is straightforward. Critics say that permanently storing images and tokens on Bitcoin imposes costs on every node operator forever, competes with payments for block space, and drifts away from what Bitcoin is for. Supporters say Bitcoin should be neutral toward any valid transaction that pays the fee, and that deciding which transactions deserve block space is a form of censorship the network was built to resist.
Bitcoin Core v30, October 2025. Bitcoin has an opcode called OP_RETURN that lets a transaction carry a small data payload. It had a default relay limit of 83 bytes. Core v30, released on October 10, 2025, raised that default to 100,000 bytes, which removes any practical constraint. The reasoning was that the cap had failed on its own terms. Ordinals had simply routed around it by using witness data instead, and the workaround produced worse outcomes for node operators than the thing the cap was meant to prevent. Release notes for every version are published at bitcoincore.org. Users can still set their own limit with -datacarriersize.
Not everyone accepted that reasoning. Bitcoin Knots, an alternative implementation maintained by Luke Dashjr with stricter relay defaults, grew from around 1% of reachable nodes in early 2024 to somewhere between 8% and 23% in 2026, depending on which measurement you trust. Knots never rebased onto the v30 codebase.
BIP-110, August 2026. Relay policy only decides which transactions a node passes on to its peers. It is a preference, not a rule, and miners are free to ignore it. A miner could take an inscription directly and put it in a block whether or not any particular node would have relayed it, which is why settings like -datacarriersize were never going to stop inscriptions on their own. So the next move was to propose restricting arbitrary data through consensus rules, which every node must enforce. It began as BIP-444 in October 2025, became BIP-110, known as the Reduced Data Temporary Softfork, and reached Complete status in the Bitcoin BIPs repository on June 25, 2026. It layered seven new consensus restrictions, aimed squarely at inscriptions, BRC-20 and Runes, for a trial period of about one year.
It did not go well. Supporters lowered the activation threshold to 55%, well below Bitcoin's historical 95% bar, and even that proved far out of reach. Miner signaling peaked at 2.53%. Michael Saylor published a 110-point essay opposing it in July 2026, arguing consensus rules should address demonstrated security threats rather than judgements about a transaction's purpose.
Because BIP-110 was deployed as a user-activated soft fork, its mandatory signaling window opened at block 961,632 on August 8, 2026, at which point enforcing nodes began rejecting any block that did not signal support. Almost none did. Those nodes split onto their own branch, which produced two blocks in roughly eight hours and then effectively stalled with an estimated 0.15% of hash rate behind it while the main chain raced ahead. As of early September 2026 that branch remains a minority curiosity, no major exchange has listed it, and Ordinals, inscriptions, BRC-20 and Runes continue to operate on the main Bitcoin chain exactly as before.
Two things are worth taking from this. Bitcoin's consensus rules are extremely hard to change without broad agreement, which is a feature. And a consensus change could never have removed existing inscriptions anyway, because rules apply to future blocks, not past ones. Everything already inscribed is there permanently.
The underlying disagreement is not settled. It has simply returned to being an argument rather than a fork.
Where Ordinals Stand Today
Bitcoin Ordinals are a numbering system for satoshis plus a way to inscribe content onto them, producing digital artifacts stored entirely on Bitcoin. Nearly four years after the first inscription, the protocol has produced more than 90 million of them and permanently altered how Bitcoin's block space is used, even as trading volumes have fallen far from their 2024 peak and an attempt to restrict the practice at consensus level collapsed in August 2026.
If you are considering collecting, the technology is stable and the risks are well understood, but the market is small, illiquid, and dependent on infrastructure that has already proven it can disappear. Learn how UTXOs work before you buy anything.
