Starting with the arithmetic, not the adjective
One market-research forecast — cited here by name and treated as one input, not settled fact — estimates the AI-semiconductor sub-segment at roughly $65–103 billion in 2025, projecting it to exceed $1.1 trillion by 2035 at a 27–33% compound annual growth rate [1]. Taking the midpoint of that 2025 range, roughly $84 billion, against $1.1 trillion a decade later implies the market would need to grow by a factor of roughly 13x over ten years. This briefing is about what that specific number implies, not about repeating it as a promotional talking point.
Why “13x over a decade” is a meaningfully different claim than “the boom continues”
A market that merely continues growing steadily looks very different, arithmetically, from one that is forecast to grow roughly thirteenfold. The latter implies that whatever has already been built — and this cohort’s other briefings document a great deal already built, including $725 billion in combined 2026 hyperscaler capex alone [3] — represents, on this specific forecast’s own terms, a comparatively small fraction of the eventual scale being projected. That is the literal arithmetic behind this briefing’s title: if the forecast is roughly right, today’s position is closer to the foot of the growth curve than its middle or its peak.
Where this framing has to be careful not to overreach
This briefing is explicitly not arguing the forecast is correct, nor is it arguing the current AI infrastructure buildout is risk-free or guaranteed to continue at anything like a 27–33% CAGR for a full decade — this cohort’s companion bubble-debate briefing covers the case for caution directly, and any forecast projecting a sustained double-digit CAGR for ten consecutive years carries substantial execution and demand-realization risk that a single cited number cannot capture. What this briefing is arguing is narrower and more defensible: the specific, attributed 27–33% CAGR forecast, if treated as a working assumption rather than a certainty, implies an “early stage” characterization that is arithmetically grounded rather than merely a rhetorical flourish borrowed from unrelated industry-hype cycles.
How this compares to the whole-industry growth curve
It is worth contrasting this AI-specific trajectory against the much more conservative whole-semiconductor-industry forecast covered in this cohort’s companion “five analysts” briefing: one estimate puts total semiconductor revenue, across every category, at roughly $1.27 trillion by 2035 — implying a 7.36% CAGR from a roughly $1.3 trillion 2026 base [4]. The AI-specific sub-segment’s projected 27–33% CAGR is three to four times steeper than that whole- industry baseline, which is the actual quantitative basis for treating AI semiconductors as a distinct, faster-growing sub-cycle within the broader industry rather than simply riding the industry’s average growth rate.
The honest caveat this framing depends on
Every claim in this briefing depends on one specific forecaster’s model holding up over a full decade — a genuinely long horizon in an industry this cohort’s other briefings document as prone to sharp cyclical corrections, geopolitical shocks, and technology-transition risk. The “foot of the mountain” framing is a defensible reading of the cited numbers as they stand today, not a prediction this briefing is making with certainty. Readers using this framing elsewhere should carry that same caveat forward: it is early-stage relative to one specific, attributed growth forecast, not a claim that continued expansion at this pace is assured.
What history says about ten-year compounding forecasts specifically
This cohort’s history track documents at least one directly relevant precedent worth weighing against any ten-year forecast: Moore’s Law itself held for roughly six decades not because compounding growth was ever physically guaranteed, but because an entire industry organized its capital spending and engineering roadmaps around treating the target as achievable. A comparable dynamic could plausibly apply here — a widely cited 27–33% CAGR forecast becoming partly self- reinforcing if enough capital allocators treat it as their planning baseline — but that mechanism cuts both ways, since a forecast can also lose its self-reinforcing power quickly if early results disappoint and capital reallocates elsewhere, a risk this briefing’s arithmetic-only framing does not, by itself, rule out.
The version of this argument that holds up without the specific number
Even a reader skeptical of the specific 13x multiple can accept a weaker, more defensible version of this briefing’s argument: relative to the scale of AI infrastructure buildout most forecasters, across the wide range covered in this cohort’s companion “five analysts” briefing, expect over the next decade, the hyperscaler capex and accelerator shipments already documented throughout this cohort represent an early tranche of a much longer buildout cycle rather than its culmination. That weaker claim survives even if the specific $1.1 trillion figure proves too high or too low in hindsight, which is precisely why this briefing treats the number as illustrative arithmetic rather than as a load-bearing prediction on its own.
