The fee market prices inclusion. It does not price permanence. The gap between the two is the externality.
The fee market measures one thing: inclusion. A writer attaches a fee denominated in sat/vB. The miner includes the transaction if the fee is high enough. The fee market prices the act of getting into the block.
This is the scarce resource the fee market sees. Block space is limited. Demand for block space varies. The fee market clears the market: when demand is high, fees rise. When demand is low, fees fall. This is standard price theory. It works for what it measures.
What it does not measure is persistence. Once a transaction is in the block, the network stores it forever. Every node on Earth accepts, validates, and serves those bytes for the lifetime of the chain. The cost of persistence is real: disk, bandwidth, validation CPU, RAM. But the fee market does not price it. The fee is a one-time payment for inclusion. The cost of storage is perpetual.
In a well-functioning market, prices reflect costs. A producer who pays more to make a product charges more to sell it. The price signal tells the consumer what the product actually costs. The consumer decides whether to buy.
Bitcoin’s fee market is missing a price signal. The cost of writing a byte to the chain is not the same as the cost of storing that byte forever. Writing is a one-time cost: the miner’s energy to include the transaction. Storage is a perpetual cost: every node operator who carries the chain forward, forever. The fee market prices the first. It is silent on the second.
This is a market failure. The price signal does not reflect the full cost. The writer sees a fee. The network bears a larger cost. The gap between the two is the externality. It is paid by node runners. It is paid in perpetuity. It is the hidden subsidy at the heart of the settlement layer.
The fee market was already missing the persistence price signal. The witness discount made it worse.
A witness byte costs 1/4 the fee of a non-witness byte. But it costs the same to store. The discount is applied to the price signal. It is not applied to the object. A byte on disk is a byte on disk, regardless of whether it came from the witness or the non-witness portion of the block.
The discount was designed to reflect a real cost difference: witness data was supposed to be prunable. Spent outputs could discard their witness data. The discount reflected the lower storage cost of prunable data. But inscriptions made witness data permanent. Outputs were designed to never be spent. The prunability assumption collapsed. The discount remained.
The result: permanent data is priced at 1/4 the cost of permanent data. The externality is 4× larger than it needs to be. The node runners pay the difference.
Every byte the chain carries has a real cost. Not a theoretical cost. A real cost.
Disk: every full node stores the entire chain on disk. The chain grows by roughly 1 MB per block. At one block per 10 minutes, that is roughly 52 GB per year. Every year, every full node operator adds 52 GB of storage to keep up. This cost is not optional. This is what it means to run a node.
Bandwidth: every full node transmits the chain to peers. New nodes syncing the chain must download every block from genesis. The larger the chain, the longer the sync, the more bandwidth consumed. This cost grows with every block.
Validation: every full node validates every transaction in every block. The more data in a block, the more data to validate. The more complex the scripts, the more CPU cycles to execute. This cost is per-block, per-node, forever.
RAM: every full node keeps the UTXO set in memory for fast validation. The UTXO set grows with every unspent output. The more data in the chain, the larger the UTXO set, the more RAM required. This cost is ongoing and growing.
These costs are real. They are paid by node runners. They are paid forever. The fee market does not price them. The witness discount makes them worse. The gap between what the writer pays and what the network costs is the externality.
The fee market has an asymmetry. The writer pays once. The network pays forever.
A writer attaches a fee and broadcasts a transaction. The miner includes it. The block is found. The fee is paid. The transaction is confirmed. The writer’s obligation is complete.
The network’s obligation has just begun. Every node on Earth must now store, validate, and serve those bytes for the lifetime of the chain. The cost is not a one-time cost. It is a perpetual cost. It is paid by every node operator who has ever run Bitcoin Core, present and future, in perpetuity.
The fee market prices the writer’s obligation. It is silent on the network’s obligation. The asymmetry is the externality. The writer pays for inclusion. The network pays for permanence. The gap between the two is the hidden subsidy.
The diagnosis is clear. The fee market prices inclusion. It does not price permanence. The gap between the two is the externality. The witness discount amplifies it. The node runners pay it.
The fix is less clear. 1 bit = 1 bit is the simplest approach: remove the witness discount, let all bytes cost the same fee. This removes the artificial subsidy. It does not fully price permanence. But it makes the externality visible.
A complete solution would price permanence explicitly. But permanence is hard to price. The cost of storage changes over time. The network’s capacity changes over time. The fee market cannot know the future cost of storing a byte forever. It can only know the present cost of getting into a block.
The open question: can the fee market adapt to price permanence without a protocol change? Or does the protocol need to explicitly price the cost of forever? The answer is not obvious. The fee market has adapted to every scarcity it has encountered. But permanence is a different kind of scarcity. It is not a one-time cost. It is a perpetual cost. The fee market may not be designed to price perpetual costs.
What is clear is that the current system does not work. The fee market prices inclusion. The network pays for permanence. The gap between the two is the externality. The node runners pay it. The writers do not. This is not a market. It is a subsidy.