AI Memory Verdict · Aggregator lens · 2026-W19

Everyone agrees + deep supply → healthy

HOLDSTABLEAI ≈ OEM language· cos 0.88partial · missing DIST
▸ Panel signature · what AI saw, what we compared against
AI: 7 textsOEM: 2DIST: 0AGG: 1SEARCH: 10
cohort disagreement: 25.0%
1024-dimensional centroid embeddings · all numbers reproducible from methodology

Aggregator action: category vocabulary consolidating · default surface OK.

memory· level 1· tier A· 8 children· 44,349 SKUs measured· 192 observations · 5 vantagesavailable

Where you rank in Memory

34 brands · 2026-W19

For each brand AI associates with this category, we record the rank (by AI-claim count) and which reference frame AI uses when talking about the brand. When the brand ahead of you appears in a different column than your marketing speaks, that's the campaign brief.

Category leaderboard · top 3
#1
micron.com
AI ≈ OEM
n=12
#2
skhynix.com
AI ≈ OEM
n=11
#3
vikingtechnology.com
AI ≈ OEM
n=10

AI ≈ OEM

34 brands

AI talks about these brands in OEM-side terms.slug cos 0.88

  • #15
    ti.com
    n=3
    3/4 vantages
  • #8
    renesas.com
    n=9
    4/4 vantages
  • #3
    vikingtechnology.com
    n=10
    2/4 vantages
  • #18
    optifuse.com
    n=2
    2/4 vantages
  • #16
    onsemi.com
    n=3
    2/4 vantages
  • #3
    microchip.com
    n=10
    2/4 vantages

AI ≈ DIST

0 brands

AI talks about these brands in distributor-catalog terms.

No brand AI-anchored to this frame for this category.

AI ≈ AGG

0 brands

AI talks about these brands in aggregator-discovery terms.

No brand AI-anchored to this frame for this category.

AI ≈ SEARCH

0 brands

AI talks about these brands in buyer-search terms.

No brand AI-anchored to this frame for this category.
rank = AI-claim count desc within the category · top 6 brands per reference frame · week 2026-W19
Buyer-search vocabulary · week 2026-W19
source · Google Keyword Planner data

What buyers actually search for — Memory

AI's vocabulary for this category does not currently match buyer-search. These are the buyer queries — the language AI is not yet echoing.

buyer querymonthly searchescompetition
psychology673KLOW
basics of psychology301KLOW
philosophy definition61KLOW
eidetic memory61KLOW
intellect definition50KLOW
intellectuality definition50KLOW
intellectual definition50KLOW
definition for intellectual50KLOW
cognizance definition50KLOW
intellectualism definition50KLOW
ranked desc by monthly searches · seed term dropped (always #1, uninteresting) · top 10 of 10 pulled · total monthly volume shown: 1.39M
peak: psychology @ 673K searches/month · methodology and full caveats at /methodology
Atoms × Bits · 4th-axis lens
source · SEC 10-K filings · cohort tier-1

Memory: AI routes attention. Physics routes atoms.

We measure both. Below — what AI hears about this category, and what the brands themselves told the SEC about where it is actually made.

Semantic axis · AI memory verdict
AI ≈ OEM language· cos 0.88
quadrant: HOLD
From AI's vocabulary alone, this category reads as a settled, channel-coherent name.
Location axis · supply-chain dependency
TSMC dominates wafer supply.
3 of 3 disclosing brands name TSMC.
Sits on the Taiwan Strait· baseline risk 85/100. Maritime passage between Taiwan and Fujian China; ~50% of global container traffic transits adjacent waters
What this means · Memory

TSMC dependency in this cohort is partial (~43% of disclosing brands name TSMC as primary), shared with Samsung / SK Hynix / Micron in-house fabs. Risk is concentrated in advanced-node logic die (controllers, HBM stack logic) rather than commodity DRAM/NAND cells; impact reads as quality / latency rather than total supply loss.

Receipts — the brands' own SEC filings
broadcom.comAVGOSingle-source dependency
10-K · 2025-12-18
During fiscal year 2025, approximately 95% of the wafers manufactured by our CMs were produced by TSMC.
intel.comINTCSingle-source dependency
10-K · 2026-01-23
Some of our most advanced current and future products are or will be either exclusively manufactured by TSMC or reliant upon critical components, including various compute die, manufactured by TSMC.
nxp.comNXPICapacity contention
10-K · 2026-02-19
Our SSMC wafer fab facility, which produces 8 inch wafers, is jointly owned by TSMC and ourselves.
AI memory cannot read a 10-K supply-chain disclosure. The category appears healthy in AI's vocabulary; the disclosing brands themselves describe a single-point-of-failure on a known chokepoint. Two axes, one truth.

Sub-categories (8)

What you unlock

Memory — paid-tier surface

🔒 paid tier
63 brands competing here
Per-brand convergence scores, vantage coverage, full sourcing-card drill-down per brand.
7 brand category peers identified
Qwen-grounded peers with rationale, confidence, and verify_targets per peer.
192 manufacturing facilities · 42 brands · 26 countries
Geographic concentration, top brands by facility count, facility-type mix, supply-risk story.
Live week-over-week trajectory
Fragmentation direction, channel-pull attribution, drift narratives.
Anonymous visitors and crawlers see this preview. Sign in or subscribe to unlock the memory data layer.
Sign in
Public preview · counts only · no brand names, scores, or specific facility data shown
substrate · 21 weeks · 26.4M AI observations · 35,746 brands · 2025-W512026-W20
this page · 2026-W19 · numbers reproducible from our weekly substrate (methodology)
per-(brand × vantage × form) aggregation, top-form share within (brand × vantage) cells; divergence classified at category level using a 1024-dimensional embedding centroid.