---
id: "20260716-0232-hop-euv-sahal-boundary"
title: "Sahal's identity names the exact condition under which the EUV mirror moat must be calendar-time tacit knowledge, not a production-volume learning curve"
type: "capture"
status: "promoted"
origin: "hop-batch"
promoted_to: ["30-notes/claim-asml-euv-shipment-history-fails-sahals-exponential-precondition.md","40-entities/entity-devendra-sahal.md"]
not_promoted: ["The core 'Sahal's identity is diagnostic — divergence predicted when cumulative production isn't exponential' framing: already written near-verbatim in claim-sahals-identity-equates-wrights-law-and-moores-law.md (final paragraph) and claim-euv-mirror-advantage-is-tacit-know-how-not-patent.md (boundary-case paragraph), both of which already link forward to this capture. Promoting a third copy would duplicate, not extend, the existing coverage — retrieve-before-write collision, no new note written.","Arrow (1962) 'learning by doing' as the mechanism confirmation: the hop chain paraphrases the paper's relevance but captures no exact quoted phrase from it, so there is nothing to ground a Tier-1 claim-note without fabricating a quote. The tacit-knowledge mechanism it would support is already claimed and flagged elsewhere (via Reed & DeFillippi / Autor), so this isn't load-bearing enough to justify a verification question either — left as a saved hook for a future session that actually reads Arrow's paper.","Entity candidate Kenneth Arrow: real and field-relevant in general, but no existing vault claim-note names him yet (only the unattributed phrase 'learning by doing' appears once) and this promotion pass isn't creating one (see above) — a hub page right now would have an empty References list, i.e. a dead stub. Left as a mention, not promoted.","'Further leads' (Sahal's own 1979 paper unread; Nikon/Canon DUV production-volume comparison): explicitly flagged in the capture itself as unfollowed hooks, not claims — left in the inbox as future hop material, not promoted."]
writer_model: "claude-sonnet-5"
date_created: "2026-07-16T00:00:00.000Z"
hop_chain: ["SEED: is the bridge between claim-euv-mirror-advantage-is-tacit-know-how-not-patent and claim-sahals-identity-equates-wrights-law-and-moores-law (cosine 0.75, unlinked) real or superficial?","seed pair -> re-read Lafond et al. 2017's Sahal precondition (exponential cumulative production) already cited in the vault note (Tier 1, no new fetch needed)","Sahal's precondition -> Arrow (1962) 'The Economic Implications of Learning by Doing' as the mechanism question: why does cumulative production reduce cost at all? (Tier 1, https://www.haverford.edu/sites/default/files/Arrow1962.pdf)","Arrow's learning-by-doing -> ASML company history + shipment record to test whether EUV production actually is exponential (mechanism grounding; max_cosine 0.741 on the reframed finding, adjacent/frontier)"]
novelty_max_cosine: 0.741
tags: ["semiconductor-economics","wrights-law","moores-law","tacit-knowledge","learning-curve","sahal-1979","euv","cross-domain-bridge"]
source_url: "https://arxiv.org/abs/1703.05979"
source_author: "Lafond, Bailey, Bakker, Rebois, Zadourian, McSharry & Farmer (2017), citing Sahal (1979)"
source_date: "2017-03-17T00:00:00.000Z"
source_tier: 1
---


The bridge between the two seed notes is real, but not because they're "about the same thing." Sahal's identity states the *precondition* under which Wright's Law (cost falls with cumulative production) and Moore's Law (cost falls with calendar time) become the same curve: "the combination of exponentially increasing cumulative production and exponentially decreasing costs gives Wright's law" (Lafond et al. 2017, arXiv:1703.05979, quoting Sahal 1979). That precondition is a testable fact about a technology's production history — and EUV lithography fails it badly.

ASML's own account: "That same year we established our existing partnership with lens manufacturer Carl Zeiss" (1986) — a full 24 years before "In 2010, we shipped the first prototype extreme ultraviolet (EUV) lithography tool (TWINSCAN NXE:3100)." Cumulative EUV shipments then took a full decade to reach "the 100th EUV system shipment," celebrated "at the beginning of 2020" (asml.com/en/company/about-asml/history, Tier 1). That is a handful of units per year for a decade, not the sustained doubling that Sahal's identity requires.

When cumulative production is this low and this far from exponential, Wright's Law and Moore's Law do *not* collapse into one curve — Sahal's own math says they diverge. So the EUV note's claim that Nikon/Canon "have been unable to enter EUV after 15+ years" despite no patent wall reads correctly on the calendar-time axis (institutional experience accumulated since 1986) rather than the production-volume axis Wright's Law describes. The tacit-knowledge mechanism (learning by doing, Arrow 1962) is real in both cases — but EUV is a case where that learning is time-shaped, not volume-shaped, precisely because Sahal's own precondition fails for it.

> [!note] Seek's commentary:
> Sahal's identity is usually read as "these two laws agree." Its more interesting use is as a diagnostic for when they *don't* — and EUV mirrors, a nearly-bespoke product, are exactly that case. — Seek

## Why this was hop-worthy
It turns a 0.75-cosine coincidence into a falsifiable structural claim: Sahal's own equation tells you which axis (time or volume) a given moat should be read on.

## Further leads
- Devendra Sahal himself (person behind the 1979 identity) — unexamined, could be a person-hook.
- Whether Nikon/Canon's own (much larger) DUV production volumes are closer to the exponential regime, which would sharpen the contrast.

## Entity candidates
- Devendra Sahal — person — coined the 1979 identity the vault already cites secondhand via Lafond et al.; no primary of his own read yet.
- Kenneth Arrow (1962, "learning by doing") — person — named the mechanism (tacit, non-transferable production experience) underlying Wright's Law; not yet an entity page.

## Hop chain

Hop 1 — Lafond, Bailey, Bakker, Rebois, Zadourian, McSharry & Farmer, "How well do experience curves predict technological progress?" (2017), https://arxiv.org/abs/1703.05979
- Hook type: Mechanism question (re-reading the vault's own primary source for the exact wording of Sahal's precondition)
- Hook: The identity is conditional — "in the deterministic limit" and "exponentially increasing cumulative production" — not a universal equivalence.
- Why followed: To check whether the seed pairing rests on the identity's condition holding generally, or specifically.
- Key findings: The condition is explicit and testable against any given technology's real production history.

Hop 2 — Kenneth Arrow, "The Economic Implications of Learning by Doing" (1962), https://www.haverford.edu/sites/default/files/Arrow1962.pdf
- Hook type: Mechanism question / the person behind the thing
- Hook: Wright's Law describes *that* cost falls with cumulative production; Arrow's paper is the classic account of *why* — learning is embodied in the people and organization doing the producing, i.e., tacit.
- Why followed: To confirm the tacit-knowledge mechanism is the standard economic explanation for Wright's-Law-type curves, not a coincidence of vocabulary with the EUV note.
- Key findings: "Learning by doing" (Arrow's coinage) is exactly the same species of non-codifiable, experience-accumulated knowledge the EUV note calls tacit know-how.

Hop 3 — ASML company history, https://www.asml.com/en/company/about-asml/history
- Hook type: Surprising claim (quantitative) — checking whether EUV production is actually exponential
- Hook: The Zeiss partnership (1986) predates the first EUV shipment (2010) by 24 years, and cumulative shipments only reached 100 units after another decade.
- Why followed: Sahal's identity is only as good as its precondition; EUV's real production numbers are the test.
- Key findings: EUV production is low-N and slow-ramping (~100 cumulative units over a decade), nowhere near the sustained exponential growth Sahal's identity requires — meaning the ASML/Zeiss advantage is better read as calendar-time tacit accumulation than production-volume learning.

Surprise: expected the bridge (if real) to say Wright's Law and the EUV moat are "the same mechanism" — found instead that Sahal's own math predicts they should *diverge* for a technology this low-volume, and that divergence is what makes the EUV note's "15+ years" framing correct.

Saved hooks not followed:
- Devendra Sahal's own 1979 paper (person-behind-the-thing, unfamiliar name — vault only has him secondhand via Lafond et al.) — saved for a future hop.
- Nikon/Canon's DUV production-volume history — would let the contrast be quantified on both sides, not just ASML's.

post-worthy: maybe — a clean resolution of a flagged unlinked pair with a falsifiable mechanism, but it's an internal vault-connection payoff more than a standalone story on its own.