What Moore actually wrote in 1965
Gordon Moore, then at Fairchild Semiconductor and later a co-founder of Intel, observed in 1965 that the number of transistors that could be economically placed on an integrated circuit was doubling at a roughly regular cadence [1]. That is what Moore actually published: an empirical observation drawn from a handful of years of data, not a claim about physical law, and not originally framed as a multi-decade prediction at all [2].
Why “law” is the wrong word, and why it stuck anyway
Physical laws describe what the universe necessarily does. Moore’s Law describes what a global industry chose, collectively and repeatedly, to keep achieving — a crucial difference this briefing’s title makes plainly. The observation held for as long as it did substantially because semiconductor companies, starting with Moore’s own Intel, treated it as a target to engineer toward rather than a passive forecast to wait and see about. Research and development budgets, equipment roadmaps, and entire companies’ competitive strategies were built around hitting the next doubling on schedule [3].
The self-fulfilling-prophecy mechanism, explained
Once major chipmakers began publicly aligning their own roadmaps to the doubling cadence Moore described, competitors had strong commercial incentive to match that pace or fall behind — a coordination effect that made the industry-wide cadence more, not less, likely to hold, regardless of whether any individual year’s underlying physics made the next doubling easy or hard. This is precisely the same dynamic this cohort’s TSMC briefing documents in the present tense: an “unprecedented” N2 ramp across five fab phases in a single year is not a passive physical inevitability, it is a deliberately engineered target the company committed enormous capital toward hitting [4].
Why the “law” framing eventually becomes a liability
Treating an engineering target as though it were a law of nature creates a specific risk: it implies continuation is guaranteed, discouraging the kind of contingency planning a genuinely uncertain physical limit would warrant. This cohort’s companion briefing on what actually ends Moore’s Law covers the real physical limits — atomic-scale effects, quantum tunneling, backside power delivery — now approaching at the sub-2nm nodes covered throughout this cohort’s foundry track, where “just engineer harder” is running into constraints the original observation never had to contend with.
The honest legacy
Moore’s Law’s real legacy is not a law at all — it is proof that an entire global industry can coordinate around a shared, aggressive technical target for six decades, sustaining a pace of progress that, absent that coordination, might well have proceeded far more slowly and unevenly. That coordination mechanism, more than any specific transistor count, is the actual historical achievement worth understanding correctly.
Why the coordination mechanism itself is now under strain
This cohort’s foundry-track briefings document a shrinking number of companies still capable of executing at the leading edge — TSMC, Samsung, and Intel, with Rapidus attempting to join them. Moore’s Law-style industry coordination worked, in part, because enough independent companies were racing toward the same target that no single company’s slowdown could stall the whole industry’s progress. With the leading edge now concentrated among so few capable players, the coordination mechanism this briefing describes has itself become more fragile — a structural change worth factoring into any expectation about whether transistor scaling continues at anything like its historical cadence over the coming decade.
What replaces “Moore’s Law” as the operative industry target
Even as raw transistor-density scaling slows, this cohort’s chiplet-economics and advanced- packaging briefings document the industry shifting its aggressive coordination toward a different metric entirely: total system performance achieved through heterogeneous integration, rather than transistor count on a single die. In that sense, Moore’s Law’s real successor is not a single new observation with its own catchy name, but a broader shift in what the industry has collectively agreed to treat as the target worth racing toward next — the same underlying coordination mechanism Moore’s original observation set in motion sixty years ago, still doing its work under a different name.
A closing note on how to read the phrase going forward
Every time this cohort’s other briefings use the phrase “Moore’s Law” in passing — and several of them do, when discussing node shrinks, packaging economics, or scaling limits — it is worth mentally substituting the more accurate phrase this briefing has argued for throughout: not a law the universe enforces, but a target an industry chose, funded, and organized itself around achieving. That substitution changes very little about the specific technical claims those other briefings make, but it changes a great deal about how confidently a reader should extrapolate those claims forward, since a chosen target can be abandoned, slowed, or redirected in ways a genuine physical law never could be.
