The humanoid "Brain" is now 79.5% of a $40.7 trillion universe — and almost none of that value is a humanoid bet

An AI and memory super-cycle has turned the humanoid "Brain" into the most expensive — and least humanoid — part of the value chain. Across InfraMosaic's 255-company universe, the Brain carries $32.4tn of live value — 79.5% of the $40.7tn total — led by NVIDIA at ~$4.97tn, Micron at ~$1.11tn (52-week range $103 → $1,089) and SK Hynix at ~$1.01tn. The 147-name Body is worth just $3.24tn. Buying the Brain for humanoid exposure is now mostly buying the AI trade — according to InfraMosaic's live Market-Cap & Exposure dataset.

The mechanism: a memory cycle, not a robotics cycle

When the Humanoid 100 was first struck in February 2025, memory sat in a footnote — a Brain annex behind compute and foundation models. Sixteen months later it is two of the index's twelve largest names. Micron Technology now carries a market value of $1,106.99bn against a 52-week range of $103 → $1,089; SK Hynix is at $1,005.47bn (a KRW 238k → 2,407k swing). Both re-ratings were driven by high-bandwidth-memory demand from data-center AI — not by a single humanoid joint shipping.

That is the whole point. The Brain layer — foundation models, AI compute, EDA, simulation, vision, and now memory — has already been paid for by the AI trade. Its internal concentration is the highest of the three segments (an HHI of 3,232; compute is an oligopoly), and the value sits in a handful of names whose humanoid revenue today rounds to zero. NVIDIA's humanoid exposure is real in narrative — Isaac, GR00T, the simulation stack — but its $4.97tn capitalisation is set by GPUs sold into hyperscale, not by Optimus or Figure unit economics.

Notice the disconnect: the three biggest Brain names by value — NVIDIA, Alphabet, Microsoft — alone account for $13.63tn, or 42% of the entire Brain segment. Cap-weight any "Humanoid 100" basket and your return is overwhelmingly determined by these few AI mega-caps. The index's move since February 2025 is, in InfraMosaic's read, overwhelmingly an AI-compute and memory story, not a humanoid story.

Owning the Brain for humanoid exposure is a lottery ticket stapled to an already-expensive AI stock.

Where the humanoid-levered return actually sits

Flip the database from market cap to humanoid exposure and the leaderboard inverts completely. The top of InfraMosaic's IMES table — the names whose revenue, strategic position and moat are most tied to the humanoid build-out — is entirely Body, and almost entirely small- and mid-cap. Of the 147 Body names, 129 carry caps under $50bn; the highest-conviction reads are an order of magnitude smaller than that.

Harmonic Drive Systems and Moog both score 92 on exposure — the two highest in the universe — at $3.9bn and $12.5bn of market value. That is roughly one four-hundredth of NVIDIA's capitalisation, scoring higher on the only axis that measures humanoid leverage. The Body holds 29 of the universe's High-exposure names; the Brain, despite being four-fifths of the value, holds only nine — and most of those, like NVIDIA itself, owe their score to platform narrative rather than humanoid revenue.

The bottleneck that the cap weighting points away from

This is not a stylistic preference for small caps — it is where the physical scarcity lives. InfraMosaic's supply/demand order book — robot ramp × per-robot bill-of-quantities versus assessed capacity — shows the tightest 2030 nodes sitting squarely in this small/mid Body cohort:

Every one of those bottlenecks is owned by a Body name a fraction of NVIDIA's size — the reducer, screw and actuator makers at the top of the IMES table. China's dominance here isn't broad, it's surgical: it owns the three or four parts that can't be substituted, and the dysprosium/terbium export-control regime now sits directly on the highest-criticality nodes. The country concentration across the universe is an HHI of 2,468, but on magnets and reducers it is far higher. The cap weighting — and therefore any passive "Humanoid 100" exposure — points away from exactly this scarcity.

What to watch

1. The Brain's beta decoupling. The cleanest tell that a true humanoid cycle has begun will be a Body name moving on a humanoid deployment headline while NVIDIA and Micron sit still. Until then, treat any "Humanoid 100" return as an AI-compute and memory print, and watch the memory pair — Micron and SK Hynix — as the swing factor in cap-weighted baskets.

2. The reducer and screw makers re-rating. Harmonic Drive, Nabtesco, Moog, Renishaw and the Chinese reducer/screw cohort are where the deficit-risk meets the highest exposure scores. A pricing or order-book inflection there is the first place humanoid demand becomes visible in equity terms — long before it touches a $4.97tn AI mega-cap.

3. The rare-earth chokepoint. Magnets register as a capacity glut on the order book yet near-100% China sourcing on the input — the constraint is the dysprosium/terbium feedstock, not the magnet line. Any tightening of the Dy/Tb export regime hits the highest-criticality, lowest-cap part of the chain first.

The Brain has been priced. The Body has not. The work of separating the AI trade from the humanoid trade is exactly what the live database is built to do — node by node, name by name, with every figure checkable.