Case Study 8.2: The Fee Spike of 2023 — When Ordinals Filled the Mempool

Background

In late January 2023, a software developer named Casey Rodarmor released the Ordinals protocol, a system for inscribing arbitrary data — images, text, video, even full applications — directly into Bitcoin transactions. Within months, Ordinals and the derivative BRC-20 token standard had transformed Bitcoin's fee market in ways that no one anticipated, producing some of the highest transaction fees in Bitcoin's history and igniting a fierce debate about what Bitcoin's block space should be "for."

The Ordinals fee events of 2023 provide a vivid, real-world illustration of every fee market concept discussed in Section 8.5: mempool congestion, fee rate spikes, transaction eviction, RBF and CPFP usage, and the fundamental economics of scarce block space. They also raise profound questions about Bitcoin governance, the neutrality of the fee market, and the tension between different communities of Bitcoin users.

What Are Ordinals?

The Ordinals protocol assigns a unique serial number to each individual satoshi (the smallest unit of Bitcoin, equal to 0.00000001 BTC) based on the order in which it was mined. This "ordinal theory" creates a way to track individual satoshis as they move through transactions, effectively making each satoshi distinguishable.

Building on this, Ordinals allows users to inscribe data onto a specific satoshi by embedding it in the witness section of a SegWit transaction. The inscribed data is permanently stored in the Bitcoin blockchain. Because the data lives in the SegWit witness (which receives a 75% discount on block weight), inscriptions can be up to approximately 400 KB per transaction — large enough for high-resolution images, short audio clips, or complex text documents.

The key technical insight that made Ordinals possible was the combination of two Bitcoin protocol features: the Taproot upgrade (activated in November 2021), which relaxed limits on witness data size, and SegWit's weight discount, which made storing data in the witness cheaper per effective byte than storing it in the transaction itself.

The First Wave: January-March 2023

The initial Ordinals inscriptions in late January and February 2023 were modest — pixel art images, small text files, and experimental media. The Bitcoin community's reaction ranged from curiosity to alarm. Critics argued that storing JPEGs on the blockchain was a wasteful use of scarce block space that would drive up fees for "legitimate" financial transactions. Supporters argued that the fee market is neutral — if someone is willing to pay for block space, that demand is as legitimate as any other.

During this first wave, the impact on fees was noticeable but not extreme. The daily number of inscriptions grew from a few hundred to several thousand. Average transaction fees rose from 1-3 sat/vB to 5-15 sat/vB. Blocks started filling more consistently, but the mempool did not become severely congested.

The more significant effect was on block space composition. Before Ordinals, Bitcoin blocks contained primarily financial transactions (payments, exchange movements, Lightning channel operations). Now, a growing fraction of block space was consumed by inscription data. By March 2023, inscriptions accounted for roughly 50% of daily block space usage on peak days.

The BRC-20 Explosion: April-May 2023

The real fee crisis began in late April 2023 with the emergence of BRC-20 tokens — a standard for creating fungible tokens on Bitcoin using Ordinals inscriptions. BRC-20 tokens (named as a playful nod to Ethereum's ERC-20 standard) used JSON data inscribed on satoshis to create deploy, mint, and transfer operations for custom tokens.

The BRC-20 token craze triggered a speculative frenzy. Users rushed to mint and trade tokens with names like ORDI, PEPE, and MEME. Each mint, transfer, and trade required a Bitcoin transaction with an inscription. The volume of transactions exploded.

The May 2023 Fee Spike

In the first week of May 2023, the situation reached crisis levels:

Mempool statistics (May 7-8, 2023): - Mempool size exceeded 500,000 unconfirmed transactions (compared to a typical 5,000-20,000) - Total mempool size reached approximately 450 MB (Bitcoin Core's default limit is 300 MB, meaning many nodes were evicting low-fee transactions) - The minimum fee rate to remain in the mempool rose above 20 sat/vB - Fee rates for next-block confirmation spiked above 300-500 sat/vB

Cost implications: - A standard 2-in-2-out SegWit transaction (approximately 140 vB) at 400 sat/vB cost approximately 56,000 satoshis, or roughly $15-16 USD at the time. - At peak congestion, simple Bitcoin transfers cost $30-50 in fees. - Transactions paying less than 10 sat/vB had essentially zero chance of near-term confirmation and were being evicted from many nodes' mempools.

Miner revenue: - Total daily transaction fees briefly exceeded the daily block subsidy for the first time since Bitcoin's early days. On May 8, 2023, miners earned approximately 218 BTC in fees compared to a daily subsidy of approximately 103 BTC (at 6.25 BTC per block, roughly 144 blocks per day). - Some individual blocks contained over 6 BTC in fees, more than the 6.25 BTC subsidy.

Real-World Impact

The fee spike had immediate consequences for different Bitcoin user groups:

Ordinary users attempting to make Bitcoin payments were priced out. A user wanting to send $20 worth of Bitcoin faced fees exceeding the transfer amount. Many postponed or canceled transactions.

Lightning Network operators were affected because opening and closing Lightning channels requires on-chain transactions. The cost of channel management spiked, and some Lightning routing nodes reported difficulty with force-closing channels due to high fees. Pre-signed commitment transactions with outdated fee rates became unreliable.

Exchanges that had not implemented batching or SegWit faced enormous fee bills for processing withdrawals. Some exchanges temporarily suspended Bitcoin withdrawals or imposed high withdrawal fees. Exchanges that had implemented efficient batching fared better, as the per-payment cost of a batched withdrawal is much lower.

BRC-20 speculators, ironically, were both the cause and the victims of high fees. Minting and transferring BRC-20 tokens became expensive, eating into the speculative profits that motivated the activity. This created a natural feedback loop: as fees rose, the marginal speculator was priced out, eventually reducing demand.

The Second Wave: November-December 2023

After a relative calm during the summer, a second and even more severe fee event occurred in November-December 2023. This was driven by a combination of factors:

• A resurgence in Ordinals activity, including a new wave of BRC-20 speculation
• The approach of the April 2024 halving, which increased attention on Bitcoin
• Growing adoption of other inscription-based protocols (Stamps, Atomicals, Runes precursors)

Peak statistics (December 16-17, 2023): - Next-block fee rates exceeded 700 sat/vB - A standard transaction cost over $40 in fees - The mempool contained over 400,000 unconfirmed transactions - Some low-fee transactions from weeks earlier were still unconfirmed

The December spike was particularly notable because it persisted for several weeks rather than resolving quickly. From mid-November through late December, fee rates rarely dropped below 50 sat/vB, establishing a sustained high-fee environment rather than a brief spike.

Fee Market Mechanics in Action

The Ordinals fee events illustrate several fee market dynamics discussed in this chapter:

Mempool as Priority Queue

The mempool functioned exactly as the priority queue model predicts. Transactions were ordered by fee rate, with the highest-fee-rate transactions at the top. As new high-fee inscriptions entered the mempool, they displaced lower-fee financial transactions. The mempool minimum fee rose as the pool filled beyond the 300 MB default limit, and the lowest-fee transactions were evicted.

Node operators who configured larger mempools (e.g., 1 GB) retained more low-fee transactions, but since miners typically had default-sized mempools, this provided little practical benefit — the transactions still would not be mined until fees dropped.

RBF Usage Surge

RBF usage spiked during the congestion periods. Users who had initially broadcast transactions at what seemed like reasonable fee rates found their transactions stuck as the market moved against them. The ability to bump fees via RBF was, for many users, the difference between a transaction confirming within hours versus being stuck for days.

Data from blockchain analytics firms showed that RBF replacement transactions increased from roughly 2-3% of all transactions pre-Ordinals to 8-12% during peak congestion periods. Full RBF (mempoolfullrbf) also saw increased adoption among node operators, as the need for fee bumping became acute.

CPFP as a Lifeline

For transactions that were not marked as RBF-eligible, CPFP became the primary fee-bumping mechanism. Exchanges, in particular, used CPFP extensively to accelerate incoming customer deposits that were stuck at low fee rates. The economics of CPFP meant that the recipient often bore the cost of fee bumping — a notable shift from the usual expectation that the sender sets and pays the fee.

Fee Estimation Failure

One of the most impactful consequences was the failure of fee estimation algorithms during rapid congestion changes. Fee estimators work by analyzing recent blocks and the current mempool to predict future fee rates. During the sudden onset of Ordinals-driven congestion, the mempool state changed faster than estimators could adapt. Users who relied on wallet fee estimates often found themselves with transactions that were adequate when broadcast but insufficient minutes later.

This highlighted a known limitation of fee estimation: it works well in steady-state conditions but poorly during rapid transitions. The Ordinals events prompted several wallets and services to improve their fee estimation algorithms and to more prominently offer manual fee selection.

Batching Payoff

The fee events starkly illustrated the value of transaction batching. An exchange processing 100 individual withdrawals at 200 sat/vB would pay approximately 100 * 140 vB * 200 sat/vB = 2,800,000 sats in fees. The same 100 payments batched into a single transaction (approximately 100 * 31 vB + 10 vB overhead per output, plus ~68 vB for input, roughly 3,168 vB) at 200 sat/vB would cost approximately 633,600 sats — a 77% reduction. Exchanges that had invested in batching infrastructure saved millions of dollars in fees during the congestion events.

The Governance Debate

Perhaps the most significant impact of the Ordinals fee events was the governance debate they sparked. The Bitcoin community found itself divided into several camps:

"Bitcoin is for financial transactions": This camp argued that Ordinals and BRC-20 tokens were "spam" that congested the network and priced out legitimate users. Some proposed soft-fork changes to Bitcoin Core's relay policy or even consensus rules to limit or prevent inscription transactions. Proposals included limiting the size of witness data, implementing OP_RETURN-style data anchoring instead of witness inscriptions, or adjusting the SegWit witness discount.

"The fee market is neutral": This camp argued that block space is a scarce resource allocated by fee market competition, and any willing buyer is a legitimate user. If inscription users are willing to pay for block space, their demand is as valid as any other. Attempting to censor specific transaction types would undermine Bitcoin's neutrality and set a dangerous precedent: if the community could decide that NFTs are "spam," what would prevent a future decision that privacy transactions or small payments are also "spam"?

"This is good for Bitcoin": Some argued that Ordinals demonstrated Bitcoin's versatility and, more importantly, provided a preview of how transaction fees could sustain mining security as the block subsidy continues to halve. If Bitcoin's long-term security depends on transaction fees replacing the subsidy, then any demand for block space is beneficial, regardless of its purpose.

"This is a temporary phenomenon": Pragmatists noted that speculative frenzies are self-limiting. The BRC-20 craze would burn out as speculative capital moved on, and fees would return to normal. The fee market would handle the allocation efficiently, and no protocol changes were needed. This prediction proved partially correct: fees did decline significantly after each spike, though Ordinals-related activity established a new baseline level of demand.

Outcomes and Lasting Impact

The Ordinals fee events of 2023 had several lasting consequences:

1. Fee market maturity. The events forced the broader Bitcoin ecosystem to take fee management seriously. Wallets improved fee estimation. More exchanges implemented batching and SegWit. RBF support became standard rather than optional. The mempool.space visualization tool became a critical real-time monitoring resource for the entire community.

2. Renewed interest in Layer 2. The fee spikes reinvigorated interest in the Lightning Network and other Layer 2 solutions that move transactions off-chain. Lightning capacity and channel counts increased as users sought alternatives to expensive on-chain transactions.

3. The SegWit discount debate. Critics of Ordinals pointed out that the 75% witness discount (a feature of SegWit designed to incentivize adoption) was being exploited for a purpose it was never intended for: cheap data storage. This sparked discussion about whether the witness discount should be modified, though no concrete proposals gained consensus.

4. Mining revenue validation. The events demonstrated that meaningful transaction fee revenue is possible. This partially addressed long-standing concerns about Bitcoin's security budget as the subsidy halves. However, critics noted that fee spikes driven by speculative crazes are not a reliable, sustainable revenue source.

5. The Runes protocol. Casey Rodarmor, the creator of Ordinals, developed the Runes protocol (launched at the April 2024 halving) as a more block-space-efficient alternative to BRC-20 for fungible tokens on Bitcoin. Runes was explicitly designed to reduce the per-token-operation block space footprint, partially addressing the congestion concerns.

Discussion Questions

1. The Ordinals fee events can be framed as either "spam attacks on Bitcoin" or "legitimate demand for block space." Which framing is more accurate, and why? Is there an objective criterion for distinguishing "spam" from "legitimate use" in a permissionless system?

2. During the May 2023 fee spike, a user wanting to send $20 in Bitcoin faced fees exceeding $30. What does this imply for Bitcoin's suitability as everyday payment money? Does the existence of Layer 2 solutions like Lightning fully address this concern?

3. Some have proposed modifying Bitcoin's consensus rules to limit inscription sizes or adjust the SegWit witness discount. Evaluate the risks and benefits of such a change. What precedent would it set?

4. The fee spikes were partially self-correcting: as fees rose, marginal speculators were priced out, reducing demand. Is this self-correcting mechanism sufficient, or should the protocol include additional mechanisms to smooth fee volatility?

5. Miners earned record fee revenue during the Ordinals events, with some blocks containing more fee revenue than the block subsidy. If sustainable, this would significantly improve Bitcoin's long-term security budget. But is speculative-driven fee revenue sustainable? What would a more stable fee revenue model look like?

6. Compare the Ordinals fee debate to historical debates about "appropriate" uses of the internet (in the 1990s, some argued that using internet bandwidth for streaming video or music was a waste of resources that should be reserved for "serious" academic and business use). What parallels and differences do you see?