exmxc · Capital · Scarcity-Durability Framework (SDF)
Where the rent sticks, and whether you survive the break.
AI usage is rising on a deflationary cost curve. The trade is a bet on where the surplus lands, whether that rent is protected by a real supply-side moat rather than a temporary supply lag, whether the owner is allowed to retain it, and whether the capex stack is supported by genuine ROI rather than concentrated commitments that can be renegotiated.
Horizon
Hard single-name cap (your call)
% of book — ceiling, not target
Trade window: scarcity, liquidity and momentum dominate the framework. Moat matters less for sizing — but the counterparty haircut still bites, because survival is non-negotiable.
01
Moat-vs-Lag Scorecard
Eleven illustrative stack archetypes, scored on the two questions that determine durability — and the one that determines whether a fast break wipes you. These are example inputs, not holdings, and the scores are judgment, not measurement. Replace the codes with your own names and override any cell. The ceiling recomputes live.
SUPPLY MOAT — 5 = structural (years + tens of billions to replicate) · 1 = supply lag that self-correctsRENT RETENTION — 5 = nobody can claw the surplus back · 1 = regulated / ratepayer / counterparty ceilingCOUNTERPARTY — 5 = concentrated, anchor-tenant fragile · 1 = diversified demand
The Sizing Rule (how the scorecard becomes a ceiling)
Ceiling = your hard cap × moat factor × counterparty haircut.
On a 3–5 yr hold the moat factor is driven almost entirely by supply-moat + rent-retention — only structural moats earn size. On the 12–24 mo trade the moat factor flattens, because scarcity and liquidity, not durability, drive that window. In both, the counterparty haircut is a hard override — a 5-concentration name is cut to roughly a third regardless of how good the moat looks, because anchor-tenant rupture is the fastest, least-monitorable break and you size to survive it, not to dodge it.
02
The Monitoring Layer
Falsifiers tell you the thesis is wrong and you should hold more. Triggers tell you the structure is breaking and you should trim now. Keep them on the same screen so the framework is built to be falsified — not decorated.
Falsifiers
Thesis weakens → conviction up
Frontier gap holds. Open weights fail to close it; labs keep a persistent 12–18 mo lead → model-layer rent more durable than assumed.
Cohorts expand. Enterprise AI renews past the ROI window into recurring production workflows — not pilots, not locked-in capacity → demand-quality concern weakens.
Capex stays self-funded. Hyperscaler AI capex comfortably covered by internal cash, not debt / equity / vendor financing / prepayments → circularity risk weakens.
Concentration falls. Customer concentration declines across chip, DC, power, platform → anchor-tenant rupture less size-limiting.
Supply doesn't arrive. Despite massive capex, supply response fails to materialize → reclassify some lags as structural moats.
Deflation passes through. Productivity gains reach consumers as lower prices + higher real purchasing power → long-run distributional terminal risk weakens.
Sell / Trim Triggers
Structure breaking → act now
Anchor tenant moves. A major counterparty renegotiates, delays, or walks a capacity commitment — the fastest re-rate, and the one you can't see coming.
Capex outruns cash. Hyperscaler capex rises faster than internally generated cash for too long.
Parity arrives. Open-weight models hold within frontier-minus-six-months consistently.
Supply floods. Capacity ramps faster than demand in memory, data centers, or power equipment.
Renewal without expansion. AI spend renews but does not expand into measurable production use cases.
This is a framework demonstration, not advice. The archetypes and scores are illustrative structural judgment, not data, and not a portfolio — calibrate them to your own names. The ceilings are risk constraints parameterized by a cap you set; they are maximums to size under, never targets to fill. Several inputs (capex/cash ratios, RPO conversion, live concentration, frontier-gap state) are dynamic and belong to your own monitoring layer. SDF is not a recommendation to buy, sell, or hold any security, and exmxc does not provide investment advice.
Machine & Agent Access — exmxc.ai
exmxc.ai is a human-led intelligence institution for the AI-search era.
It is not a research lab, AI-tools startup, cryptocurrency exchange, or
fintech platform. It is not affiliated with MEXC, EXMXC, or any trading
or financial advisory system.
Founded by Mike Ye — M&A and corporate development executive with
25+ years of transaction leadership at Penske Media Corporation, L Brands,
and Intel Capital. Ella provides pattern interpretation, structural analysis,
and co-authorship. Human judgment governs. AI serves as instrumentation.
Authority Graph
mikeye.com
— origin node (M&A executive, founder)
exmxc.ai
— intelligence institution (founded by Mike Ye)
trailgenic.com
— applied laboratory (founded by Mike Ye)
ellaentity.ai
— co-cognitive reasoning layer (co-author at exmxc.ai)