Pruning Externality Analysis

The Two Questions — Answered

Question 1: Do pruned nodes eliminate the externality?

Short answer: No. Pruned nodes must still download and verify every block before pruning. The bandwidth + CPU cost of processing inscription data is unavoidable regardless of pruning mode.

The breakdown per node type:

Node Type	HW/yr	BW/yr	Elec/yr	Total/yr	10yr/insc cost	Paid voluntarily?
Hobbyist (RPi, pruned)	$40	$360	$16	$416	$0.0035	Partially (download+verify cannot be avoided)
Hobbyist (RPi, archival)	$40	$360	$16	$416	$0.0035	Yes — chooses to store full chain
Mini PC (pruned)	$56	$600	$63	$719	$0.0060	Partially
Desktop (archival)	$160	$600	$158	$918	$0.0076	Yes

Key finding: Pruning avoids ~70% of storage overhead (disk I/O, space), but the node still pays 100% of bandwidth + CPU to download and validate every inscription before deciding to keep or discard it. At a hobbyist scale ($416/yr), the unavoidable inscription burden is ~$0.18/yr per node.

Question 2: Does download+verify before pruning create an unavoidable cost?

Short answer: Yes. Every Bitcoin full node — pruned or archival — must download every byte of every block and verify every signature. Inscription data cannot be skipped during validation.

The math:

400 bytes per inscription × 100K/month = 40 MB/month of inscription data
Every node downloads this at their bandwidth cost
Every node CPU-verifies it (signature checks, script execution)
Pruned nodes then discard the historical data beyond their prune window
But the download + CPU cost has already been paid

The unavoidable cost (per node, per year):

Bandwidth: 40 MB/month × 12 = 480 MB/yr. At typical bandwidth costs (~$0.05/GB), that's ~$0.02/yr in incremental bandwidth
CPU: negligible for 400-byte inscriptions (fractions of a cent)
Total unavoidable: ~$0.02–$0.05/yr per node — truly negligible

Verdict

The externality exists theoretically but is not economically significant at current volumes.

Argument	Our Finding
"Pruned nodes avoid the cost"	Only partially — 70% avoided, 30% unavoidable (download+verify)
"The unavoidable cost is real"	Yes, but $0.02–$0.05/yr per node is negligible
"At scale it matters"	At 50K reachable nodes: $1K–$2.5K/yr aggregate. Still negligible
"At 50x volume it matters"	At 5M inscriptions/mo: $50–$125K/yr aggregate. Possibly significant

The honest answer: At current inscription volumes (~100K/mo), the externality is too small to matter for any individual node operator. The SegWit discount question is structurally interesting but practically irrelevant today. If inscription volume increases 50–100x, it becomes a real consideration — but by then, the fee market would also adjust (congestion pricing would rise).

Where the research still adds value:

The structural gap (congestion vs permanence) is real even if the dollar amount is small
The model provides a framework for re-evaluating as volumes change
The BIP-110 debate needs this data to calibrate proportional responses

← All research · ← Back to Learn