| 2026-06-ai-margin-migration-P1 | 2026-06-ai-margin-migration | Through 2027-12-31, the AI application layer splits on integration: companies that own their inference and/or proprietary data expand gross margin, while pure model-wrappers (no owned inference, no proprietary data/distribution) compress toward death. Integration — not the model — is the margin discriminator. | 2026-06-08 | 70% | 2027-12-31 | open | — |
| 2026-06-ai-margin-migration-P2 | 2026-06-ai-margin-migration | Through 2027-12-31, frontier labs do **not** suffer an absolute gross-margin collapse toward marginal token cost on their premium SKUs: premium-tier demand stays price-inelastic enough that labs hold (or grow) compute margin even as headline $/token falls. Commoditisation hits the low end, not the frontier. | 2026-06-08 | 60% | 2027-12-31 | open | — |
| 2026-06-ai-margin-migration-P3 | 2026-06-ai-margin-migration | Through 2027-12-31, the naive commoditisation reading holds in full: model-API providers broadly see gross margins collapse toward marginal token cost as open weights and price wars strip the layer. **We predict this MISSES** — commoditisation bites the good-enough tier and the wrapper resellers, not the frontier labs' own premium economics (cf. P2). | 2026-06-08 | 38% | 2027-12-31 | open | — |
| 2026-06-d2d-margin-release-L1 | 2026-06-d2d-margin-release | Through 2027-06-30, the AT&T/T-Mobile/Verizon satellite JV (and its members, individually or collectively) will not acquire majority control of AST SpaceMobile; the JV–AST relationship resolves as commercial/wholesale procurement, at most with continued minority investment. | 2026-06-08 | 70% | 2027-06-30 | open | — |
| 2026-06-d2d-margin-release-L2 | 2026-06-d2d-margin-release | Within 12 months (by 2027-06-06), AST SpaceMobile remains independent (no change of control) AND signs at least one additional carrier-collective capacity arrangement — either the US JV's definitive wholesale agreement or an international analog (e.g., a multi-carrier or national-coalition deal in Europe/Asia).
Conjunctive claim: both legs must hold for a HIT. | 2026-06-08 | 55% | 2027-06-06 | open | — |
| 2026-06-d2d-margin-release-L3 | 2026-06-d2d-margin-release | Within 12 months (by 2027-06-06), the AT&T/T-Mobile/Verizon JV's definitive arrangements will be operator-neutral or multi-vendor — it will not grant AST (or any single satellite operator) exclusivity over the JV's pooled spectrum or D2D procurement. | 2026-06-08 | 72% | 2027-06-06 | open | — |
| 2026-06-d2d-margin-release-L4 | 2026-06-d2d-margin-release | Through 2027-06-30, AST SpaceMobile **equity (ASTS)** delivers a sustained, outsized re-rating attributable to its neutral-host consolidation: the stock materially outperforms a satellite/communications benchmark off the strength of L1–L3 playing out. **Expected MISS.** The structural thesis says throughput accrues to AST's constrained operating layer — it makes no promise the *equity* captures it by this horizon. The constellation is unbuilt and capital-intensive, the stock already round-tripped an all-time high, and the binding risks (launch execution, dilution) sit outside the lens. A MISS here separates "the mechanism is right" (AST is the neutral host) from "the security is the way to express it." | 2026-06-08 | 42% | 2027-06-30 | open | — |
| 2026-06-graphite-electrode-repricing-P1 | 2026-06-graphite-electrode-repricing | By 2027-12-31, GrafTech realizes a material part of its announced electrode price increase: weighted-average realized ASP rises off its Q1'26 ~breakeven floor (~$3,900/MT against $3,848/MT cash cost) to **>$4,000/MT** — i.e. >~$400/MT of the announced $600–$1,200/MT increase realized — with **positive cash margin per MT for ≥2 consecutive quarters**, and electrode-share-of-finished-steel moving off **0.74%** back toward the historical **1.1%**. The electrode is the binding, non-substitutable EAF input; once the producer stops subsidizing steelmakers, throughput rent accrues to the constraint. | 2026-06-08 | 68% | 2027-12-31 | open | — |
| 2026-06-graphite-electrode-repricing-P2 | 2026-06-graphite-electrode-repricing | Through 2027-12-31, GrafTech sustains its declared "value over volume" posture: it **does not rescind** the March-2026 price increase and **does not chase utilization back up by discounting** below cash-margin levels. Price leads; volume/utilization may stay soft or flat rather than being bought back at sub-margin ASP. This is the causal lever behind P1 — the producer voluntarily withholds sub-margin capacity, manufacturing the very constraint that lets throughput rent accrue. | 2026-06-08 | 63% | 2027-12-31 | open | — |
| 2026-06-graphite-electrode-repricing-P3 | 2026-06-graphite-electrode-repricing | By 2027-12-31, even if the electrode reprices (P1) and discipline holds (P2), GrafTech **equity (EAF)** delivers an outsized constraint-driven re-rating: the stock materially outperforms EAF-steel / industrial peers on the strength of the margin inflection. **Expected MISS.** The throughput rent accrues to the *constraint margin*, but the equity is encumbered by the debt-service / liquidity overhang the Street is actually pricing — so the mechanism being right (margin re-rates) need not make the trade right (equity captures it). Misses here separate "the lens is correct" from "the security is the way to express it." | 2026-06-08 | 40% | 2027-12-31 | open | — |
| 2026-06-power-binding-constraint-P1 | 2026-06-power-binding-constraint | Through 2027-12-31, the capacity-capped electrical-equipment tier (gas turbines, large power transformers, HV switchgear) sustains its scarcity rent: order backlogs, lead times, and order pricing stay at or above their 2026 peak. The durable AI-power moat is the equipment slot, not the utility re-rating. | 2026-06-08 | 78% | 2027-12-31 | open | — |
| 2026-06-power-binding-constraint-P2 | 2026-06-power-binding-constraint | Through 2027-12-31, the constraint does not pass cleanly from silicon to power: GPU/accelerator supply stays co-scarce (HBM + advanced packaging allocated), so GPU gross margins do **not** visibly compress on a power-gated story. Power and silicon are *co*-bottlenecks, not a relay. (Contrarian to the "power is now THE constraint, so GPUs de-rate" reading.) | 2026-06-08 | 62% | 2027-12-31 | open | — |
| 2026-06-power-binding-constraint-P3 | 2026-06-power-binding-constraint | Through 2027-12-31, the merchant generators most associated with the AI-power trade (Vistra, Constellation, Talen) deliver *further* outsized, constraint-driven outperformance versus the broad market — i.e. the easy re-rating still has a second leg. **We predict this MISSES**: the constraint is real but already priced into these names. | 2026-06-08 | 38% | 2027-12-31 | open | — |
| 2026-06-d2d-margin-release-L5 | 2026-06-d2d-margin-release | Through 2027-06-30, the device integrator does not capture the margin the carriers released by pooling. Concretely: Apple neither (a) acquires or takes control of a satellite-constellation operator, nor (b) launches a paid consumer direct-to-device (D2D) **data** service — beyond emergency/SOS messaging — that bypasses the carrier-wholesale channel. Mainstream consumer D2D connectivity stays carrier-delivered, over neutral-host capacity, not device-bundled by Apple. This is the rival-claimant test: in a phase that rewards vertical integration, the device layer is the most credible challenger to the neutral-host operator slot, and we bet it does not win that slot by this horizon. | 2026-06-09 | 60% | 2027-06-30 | open | — |
| 2026-06-agentic-commerce-margin-C1 | 2026-06-agentic-commerce-margin | By 2026-11-30, the **Oct-2026 network-rules cycle** (VisaNet Business Enhancements October release, or Mastercard's equivalent announcement/rules publication) contains at least an **announced agentic-token / agent-identity compliance requirement with an effective date** — even if the effective date itself is in 2027. | 2026-06-11 | 35% | 2026-11-30 | open | — |
| 2026-06-agentic-commerce-margin-P1 | 2026-06-agentic-commerce-margin | By 2027-12-31, no model-layer assistant sustains a **native-checkout take-rate at or above 4%** (OpenAI's withdrawn Instant Checkout fee — the demand layer's high-water mark) at material, disclosed scale; demand-layer commerce economics converge toward ordinary affiliate/referral rates, and **checkout + fulfillment stay with merchants/PSPs**. The assistant becomes the discovery front door, not the toll booth. | 2026-06-11 | 79% | 2027-12-31 | open | — |
| 2026-06-agentic-commerce-margin-P2 | 2026-06-agentic-commerce-margin | Through the **Apr-2027 network-rules cycle** (publications by ~2027-06-30), at least one card network — Visa (TAP / Agentic Directory) or Mastercard (Agent Pay / Know-Your-Agent) — converts its **opt-in agentic framework into an effective-dated compliance rule** (agentic-token, agent-identity, or agent-registry requirement published in VisaNet Business Enhancements or the Mastercard rules/announcement equivalent, with an effective date). This is the interchange-style tax vector: the moment the framework stops being voluntary, the rails own the toll. | 2026-06-11 | 60% | 2027-06-30 | open | — |
| 2026-06-agentic-commerce-margin-P3 | 2026-06-agentic-commerce-margin | By 2027-12-31, the rival claimant wins: **Amazon converts agent *access* into a licensed, litigated property right** — the Ninth Circuit upholds the Perplexity/Comet preliminary injunction **and** Amazon ships a paid agent-access (licensing) program — capturing agentic-commerce margin **by refusal** rather than via open rails. **Expected MISS.** The thesis (P1/P2) says margin lands on shared rails because the standards-war assets sit with networks and merchants collectively; this control tests the strongest competing mechanism, where the decisive asset is a single dominant merchant's property right over its own front door. | 2026-06-11 | 37% | 2027-12-31 | open | — |
| 2026-06-ai-margin-migration-C1 | 2026-06-ai-margin-migration | Through 2026-09-30, neither Anthropic nor OpenAI cuts the published API list price of its frontier flagship tier by more than 10% from the 2026-06-11 baseline rates (pinned in the evidence log: Anthropic Claude Fable 5 $10/$50 per MTok in/out; OpenAI GPT-5.5 $5/$30 and GPT-5.5-Pro $30/$180). | 2026-06-11 | 72% | 2026-09-30 | open | — |
| 2026-06-d2d-margin-release-C1 | 2026-06-d2d-margin-release | At AST SpaceMobile's Q2'26 earnings (scheduled ~2026-08-10/17; horizon 2026-08-31), FY2026 revenue guidance is maintained at ≥ $150M (not cut below the existing $150–200M range) **and** AST has announced no control-changing equity transaction with the JV carriers (no agreement giving T/VZ/TMUS, individually or via the JV, >50% voting control). | 2026-06-11 | 78% | 2026-08-31 | open | — |
| 2026-06-graphite-electrode-repricing-C1 | 2026-06-graphite-electrode-repricing | At GrafTech's Q2'26 earnings (expected late July / early August 2026; horizon 2026-08-31), the March-2026 price increase stands un-rescinded, management maintains the "value over volume" posture, and realized ASP is not below the Q1'26 ~$3,900/MT level — i.e. no volume bought back at the cost of price. | 2026-06-11 | 80% | 2026-08-31 | open | — |
| 2026-06-graphite-electrode-repricing-C2 | 2026-06-graphite-electrode-repricing | At GrafTech's Q3'26 earnings (expected early November 2026; horizon 2026-11-30), realized weighted-average ASP prints **≥ $4,000/MT** — the announced increase visibly lands in the first quarter where management said ~90% of the pricing benefit begins (H2'26 contract resets). | 2026-06-11 | 65% | 2026-11-30 | open | — |
| 2026-06-power-binding-constraint-C1 | 2026-06-power-binding-constraint | At GE Vernova's Q2'26 earnings (scheduled ~2026-07-21/23; horizon 2026-08-15 for slack), the gas-power equipment position does not roll over: backlog is flat-or-up sequentially versus the ~100 GW Q1'26 level, and management reports neither declining new-order pricing nor materially shortening lead times. | 2026-06-11 | 85% | 2026-08-15 | open | — |
| 2026-06-power-binding-constraint-C2 | 2026-06-power-binding-constraint | At Nvidia's FQ2'27 earnings (scheduled 2026-08-26; horizon 2026-09-15 for slack), GAAP gross margin prints **≥ 72.0%** — no visible compression of the kind a clean GPU→power baton-pass would produce. | 2026-06-11 | 90% | 2026-09-15 | open | — |
| 2026-06-d2d-margin-release-L6 | 2026-06-d2d-margin-release | Through 2027-06-30, the integrated platforms that triggered this whole episode — SpaceX/Starlink and Amazon Leo, which bought the scarce spectrum outright and own full stacks — do not capture mainstream US consumer D2D by going *direct to consumers* and bypassing the carrier-wholesale channel. Concretely: neither launches, nor announces with a firm launch, a mainstream paid consumer direct-to-device (D2D) **data** service in the US sold direct (not via a carrier/MNO wholesale arrangement) to standard handsets at scale. Mainstream US consumer D2D stays carrier-delivered — including carrier-*resold* platform capacity, e.g. T-Mobile reselling Starlink. This is the prime-mover test: the platforms' prior spectrum land-grab is what forced the carrier JV, so the platforms (not just Apple's device layer, L5) are the most capitalized claimant that could hollow out the very channel the neutral-host thesis (L1–L3) rests on; we bet the channel holds through this horizon. | 2026-06-16 | 62% | 2027-06-30 | open | — |
| 2026-06-eu-resilience-auction-floor-P1 | 2026-06-eu-resilience-auction-floor | By 2027-12-31, across the 2026–27 national NZIA-criteria renewable auctions (Italy GSE FER-X follow-on rounds, France CRE, Germany BNetzA, Spanish registers), the EU's Net-Zero Industry Act non-price **"resilience" criterion behaves as a recurring, structural protected-demand floor** for non-Chinese PV — the protected lane keeps clearing **materially above the unrestricted reference and inside the instrument's own ~5–15% design band** (Italy's first round cleared **+17% gross / ~14.5% on LCOE**), rather than collapsing to a **≲5%** developer transfer. The recurrence is statutory — non-price criteria on **≥30% of annual auction volume, or ≥6 GW, per Member State, every year** (base Reg (EU) 2024/1735 Art. 26), applying from **2025-12-30** — so the claim is that the premium *persists across rounds and Member States*, not that one Italian auction printed it once.
**Region note — EU.** The carve-out is created by EU regulation (base Reg 2024/1735 + Implementing Regs 2025/1176 and 2025/1178) and is falsified in Member-State auction venues (GSE, CRE, BNetzA, Spanish registers) — that is where the clearing premium is observable. China is the *excluded* supplier and India/Korea/US are potential *beneficiary*-suppliers, but their home markets are not where this claim trips. Out of scope: non-EU procurement instruments (US IRA domestic-content is a separate mechanism, [`critical-minerals-floor`](../../candidates/2026-06-critical-minerals-floor.md) is the US analogue). | 2026-06-16 | 67% | 2027-12-31 | open | — |
| 2026-06-eu-resilience-auction-floor-P2 | 2026-06-eu-resilience-auction-floor | By 2027-12-31, at least one **named non-Chinese manufacturer at the binding non-China complement — PV cells/wafers or inverters (EU *or* non-China-but-non-EU, e.g. an Indian/Korean cell supplier or an inverter maker)** — discloses protected order-book or capacity **attributable to the NZIA carve-out**. The protected rent lands *deeper than module assembly*: because non-preferential rules of origin mean assembling cells into modules **"would not change the origin of the module"** (Impl. Reg 2025/1176 recital 10), module assembly alone does not confer non-Chinese status, so the resilience criterion bites at the cell/wafer + inverter stage. The market's reflex trade — "buy European *module* makers" — looks at the wrong layer.
**Region note — EU.** The criterion and the demand it creates are EU (the carve-out is falsified in EU auction outcomes and EU-facing capacity decisions); the *captor* may be non-EU (India's Waaree, a Korean cell maker) — that is the geography residue, not a scope change. The claim trips on a disclosure tied to **EU NZIA demand**, wherever the supplier is domiciled. Out of scope: capacity a firm builds for non-EU markets (US IRA-driven expansion is a different driver). | 2026-06-16 | 60% | 2027-12-31 | open | — |
| 2026-06-eu-resilience-auction-floor-P3 | 2026-06-eu-resilience-auction-floor | By 2027-12-31, the **naive read wins**: the protected demand revives **European module/cell manufacturing**, and the obvious "buy European solar manufacturers" trade pays off — a named EU module or cell maker captures the carve-out's rent and adds protected capacity at scale (reshoring, not leakage). **Expected MISS.** The thesis (P1/P2) says the rent does *not* reshore EU module assembly — it lands at the non-China complement (cells/inverters) and can leak to non-China/non-EU suppliers, because EU cell capacity is thin and distressed and module assembly doesn't confer non-Chinese origin. This control tests the strongest competing mechanism: that EU industrial policy + protected demand together restart domestic manufacturing inside the window.
**Region note — EU.** Same market as P1/P2 (the carve-out and the manufacturing it would or wouldn't revive are both EU); the control narrows the *captor* to EU-domiciled module/cell producers, the geography the thesis says loses out. Out of scope: EU *inverter* makers (SMA) capturing rent — that is a P2 HIT, not a reshoring revival of the module/cell layer this control is about. | 2026-06-16 | 22% | 2027-12-31 | open | — |
| 2026-06-d2d-spectrum-toll-P1 | 2026-06-d2d-spectrum-toll | Once SpaceX owns AWS-4/H-block and is independent of the carrier-lease framework, it keeps selling D2D wholesale through the carriers and the open spec rather than flipping to a consumer-direct over-the-top service that disintermediates them. This is a US thesis: US carriers, US-licensed satellite spectrum, the US regulator. | 2026-06-20 | 60% | 2027-11-30 | open | — |
| 2026-06-d2d-spectrum-toll-P2 | 2026-06-d2d-spectrum-toll | As 3GPP-standardized non-terrestrial-network (NTN) support diffuses into mainstream handsets over ~2027–2028, the per-device margin settles into a standards-essential modem royalty that Qualcomm and MediaTek collect — a toll orthogonal to which constellation (Starlink / AST / Amazon Leo) supplies the bits — rather than shipping as a free, un-tolled baseline feature.
This is a US-market thesis on a global supply input: the carriers, the licensed spectrum, and the regulator are US, while the modem SEP toll is collected worldwide on the handsets sold into that market. It is the topic's headline call: the released carrier-coverage rent lands on the scarce non-modular complement the open spec cannot reproduce — the device modem's standards-essential-patent position. | 2026-06-20 | 72% | 2028-12-31 | open | — |
| 2026-06-d2d-spectrum-toll-P3 | 2026-06-d2d-spectrum-toll | Through ~2027–2028, AST SpaceMobile stays an independent multi-tenant neutral host rather than getting absorbed or margin-compressed into dependence by an integrated owner-operator or the carrier joint venture's bargaining power. This is a US thesis; AST's non-US carrier partners and shareholders are out of scope as controllers of the US margin. | 2026-06-20 | 55% | 2028-12-31 | open | — |
| 2026-06-d2d-spectrum-toll-P4 | 2026-06-d2d-spectrum-toll | The naive tradable expression of the headline — "own Qualcomm (or MediaTek) to capture the standards-essential NTN modem toll" — does **not** decisively pay off by the ~2028 horizon. The toll is collected (P2's mechanism holds), but it is immaterial to a ~$10B+/quarter revenue base and absent from the priced narrative, so the structural win does not re-rate the security. We expect this one to **lose**.
This is the pinned expected-MISS control, not a caveat: it prices the *security* leg, where the headline P2 prices the *mechanism* leg. Both can be true at once. | 2026-06-20 | 25% | 2028-12-31 | open | — |
