The witness discount
SegWit weighs witness bytes at a quarter of base bytes. What that does to fees, how BTI measures it per protocol, and what Blockspace Value Theory proposes it means.
The witness discount is the rule, introduced with SegWit, that a byte in a transaction's witness counts 1 weight unit while every other byte counts 4. The first half of this page is the mechanism, from block.space's data-types reference, and the measurements BTI has taken of it. The second half is what Blockspace Value Theory proposes it means. That half is theory, and it is labelled as such.
The mechanism #
A segwit transaction carries one witness stack per input, after its inputs and outputs. The txid hashes the transaction without the witness and the wtxid hashes everything, which is what makes witness data structurally separable. Weight is stripped_size × 3 + total_size, so:
- every non-witness byte costs 4 WU;
- every witness byte costs 1 WU, whether it is a signature or a file;
- the consensus block limit is 4,000,000 WU.
The data-types reference sums it up: "witness bytes are real bytes on disk and on the wire, but count 1/4 toward the 4,000,000 WU consensus block limit." Fee markets price vbytes, so at a fee rate of f sat/vB a witness byte costs f/4 sats and a non-witness byte costs f sats.
What it does to data on Bitcoin #
The reference's channel comparison follows from the discount:
| channel | weight per data byte | sats per data byte at f sat/vB |
|---|---|---|
| taproot witness envelope | 1 WU | f/4 |
| OP_RETURN | 4 WU | f |
| bare multisig fake keys | 7.4 WU | 1.84f plus dust |
It explains why every high-volume protocol, ordinals, BRC-20, Alkanes WASM and Counterparty v11, converged on the witness envelope, and why OP_RETURN keeps small structured messages such as runestones, where its single-transaction atomicity matters more than the price.
Measured per protocol #
Because BTI keeps both raw bytes and vbytes on every row, the ratio of bytes to vbytes is the share of the discount a protocol takes. BTI's design names that ratio the witness subsidy. From the July 2026 measurement run:
| protocol | bytes : vbytes | reading |
|---|---|---|
| inscriptions | 3.374 | pays for about 30 % of what it consumes |
| stamps | 1.008 | essentially no discount, deliberately |
| runes (runestone) | 1.0000 | zero discount, to four decimals |
| Omni | 1.0142 | pre-segwit, as expected |
The measurements document reads this as "a 3.37× gap between two protocols using the same chain in the same year": stamps chose to forgo the discount for permanence, and inscriptions took all of it. A single large file shows it most plainly. In case 7 of the BTI case studies, a 3.9 MB jpeg inscription, Book A is 3,911,878 bytes against 978,204.25 vbytes, a ratio the case study puts at about 4.00: "this artifact bought raw bytes at a quarter weight".
blockspacelabs.com lists "The witness discount, by protocol" among its research projects: each token family's split of base and witness bytes, and what the discount saved it in fees. It is a described study; no results are published yet.
What BVT proposes (theory) #
This section reports Blockspace Value Theory, a living document and an untested thesis.
BVT §3.4 says version 2.0 of the theory was wrong to treat witness-borne tokens as proportionally less real blockspace. BVT proposes that the binding supply constraint is total verification burden, and that on that measure the channel does not matter: witness and OP_RETURN bytes are both downloaded, validated, propagated and archived by every node. It argues that pruning does not rescue the old view, because a pruned node has still paid the validation cost.
From this BVT concludes:
4 MB of witness data consumes 4 MB of the real kilobyte budget. 1 MB of OP_RETURN data consumes 1 MB.
and
The 4× discount does not create four times more blockspace. It sells the same scarce resource at one quarter the price through one particular door.
§3.5 calls the discount "a transfer of blockspace value" from the fee market to whoever routes data through the discounted channel, paid for by node operators. BVT's glossary defines the term in one line: "SegWit's 4× weight reduction on witness data. A price subsidy, not a capacity increase."
BVT names the open question this leaves: if the market ever prices blockspace tokens by kilobytes consumed rather than by fees paid, witness-era accumulation would be worth four times what its cost basis suggests; if it prices by fees, it is worth what was paid. BVT calls "kilobytes or fees" the central open variable in the theory. See Blockspace Value Theory.
Why BTI counts raw bytes #
BTI's design follows BVT here as a measurement rule, not as a claim about value. The books count raw bytes with no weight discounting, and vbytes are carried beside them so the discount stays visible on every row. See weight units and vbytes.
Related #
Sources #
Updated 2026-10-20 · This page as Markdown