←The mapResearch bank
Theses
Transformer lead times gate hyperscaler returns, not start datesPackage Perimeter Arbitrage: Optics Suppliers Capture Value NVIDIA Cannot RetainNAND Flash: Datacenter Crowding-Out Masked by Low Consumer PullCoWoS Binds the Buyer, Not the LandlordLong-context inference: memory constraint compresses cloud margin before GPU-poor plays feel itCo-packaged optics transfers margin to substrate integrators, not optics specialistsSilicon Wafer: Duopoly Capturing Nothing as Downstream Rents CompoundAdvanced Packaging: The Scarcity Premium Flows Upstream, Not DownAmazon: The Capex-Margin InversionMicrosoft: The High-Multiple Cloud Business Is Already SubscaleFoxconn: Margin Squeeze Masked by AI Server Revenue MixSK Hynix: The Packaging Bottleneck Inverts the Margin NarrativeGPU collateral decay gates buildout faster than physical supplyOracle: Monopsony Rent Capture Masked by Consolidated AccountingInnolight: Negative Operating Leverage Hiding in Hyperscaler ConcentrationInterconnection queue converts hyperscaler capex into generator rentEaton: Margin expansion telegraphs demand scarcity more than growth can showElectrical Steel Asymmetry: GE Vernova Captures Transformer Scarcity Without Steel ExposureConstellation: Hyperscaler Monopsony Masks Nuclear Fuel TransmissionABF substrate dual-gate arbitrage: substrate converters capture the spread the market assigns upstreamRack Power Density: The Liquid Cooling Adoption MirageArm: The Royalty Stream That Scales With Everyone's Margin but Its OwnHBM5: The Signal Integrity Bottleneck Transfers Margin to Silicon PHY DesignersShin-Etsu: Margin Expansion Masked by Commodity PVC ConsolidationSMIC: Subsidy Converts to Capacity Under Obscured Margin PressureSemicap service is an installed-base annuity, and the market still prices these names as capex cyclicalsAlphabet: The Capex Ratchet Liquidity Mispriced as Platform OptionalityPalantir: The Margin-Protected Infrastructure PlayAccelink: The Valuation Anomaly in China's Optical Chip IndigenizationTransceiver margin expansion is a mirage: buyer concentration at 1.6T resets pricing power the market prices as durableQualcomm's low capex masks an IP-only future neither consensus nor bears have pricedMurata: The Margin That Doesn't MoveApplied Optoelectronics: Hyperscaler Procurement Arbitrage Driving Negative SpreadsGlobalFoundries: Customer Concentration Masks Structural Insulation from AI Capex WhiplashSilicon capacitors expose Intel's package productivity deficitArista: Hyperscaler capex intensity conceals margin compression riskSilicon cannot emit light — indium phosphide is the chokepoint CPO relocates but cannot removeMoE inference: converted miners capture memory scarcity, hyperscalers leak itTesla: AI Pivot Insulated from Datacenter BottlenecksCustom Silicon Pays a Different MerchantThe ABF Chokepoint Does Not Pay Its Owners — Buy the Bottleneck Fails HereNVIDIA's Rent Compresses Through Software, Not SiliconThe China chokepoint moves from lithography to bonding — and the substitution trade is priced in the wrong categoryThe Upstream Is Single-Sourced and UnpricedOptical attach is set by topology, not by shipments — linear optical TAM models are mis-specifiedHBM moat vs. DRAM commoditization — is SK Hynix's HBM-mix 'miss' actually bullish?Power Caps the AI Buildout — the bottleneck moved from chips to megawattsEMIB Shifts the Packaging Bottleneck to SubstratesASML / China-DUV displacement is overpriced — the sell-off prices near-full substitution off a 5-machine data pointCost Per Token Is Set Outside the ChipChipflation Is a Wafer Allocation, Not a Demand ShockThe Neocloud Rent Is Consumed By The Asset Before It Reaches EquityMemory-demand durability — does algorithmic efficiency cap the memory super-cycle?HBM4 Qualification Concentrates Share, Not Spreads ItNeocloud margin re-rating — do open-source deployment + rising rental rates break the thin-margin cap?The Megawatt Is the Unit of Account
Mechanisms
ABF substrate and build-up film supplyCoWoS advanced-packaging capacityConventional DRAM and NAND supply (HBM crowding-out)GPU residual value as loan collateralGrid interconnection queue positionHeavy-duty gas turbine delivery slotsLarge power transformer lead timesSamsung memory long-term agreementsBuyer concentration tighteningHBF consortiumRent converting into capacityRent migrating upstreamRent not being competed awayUALink ConsortiumUltra Ethernet Consortiumco-packaged-optics displaces copperco-packaged-optics displaces optical-transceiveremib displaces cowosglass-substrate displaces abf-substratehybrid-bonding displaces euvsilicon-capacitor displaces mlccChip designDatacenter mathEfficiency arrives in steps, not trendsInference shapeMemory economicsPhotonicsTau scalingToken mathTokenomicsCapacity arriving — CoWoS (Chip-on-Wafer-on-Substrate)Capacity arriving — HBM4Capacity arriving — Silicon Wafer

Ultra Ethernet Consortium

countervailing  a named mechanism, not a conclusion

7 firms publish scale-out-network together against nvidia's nvlink. A consortium is a revealed statement about who lacks the power to impose a standard alone — nobody forms one around something they can mandate. It kills the sole-source objection and leaves the adoption question untouched. Read it against `rent-persistence`: this is what the competitive response looks like when the rent is an interface rather than a price.

Opposes rent-persistence — this is what the competitive response looks like against that mechanism.

The path

nodeeffect
Advanced Micro Devices+0.15
Arista Networks+0.15
Broadcom Inc.+0.15
Cisco Systems+0.15
Meta Platforms, Inc.+0.15
Microsoft Corporation+0.15
Oracle Corporation+0.15
NVIDIA Corporation-0.20

What would break it

The rent survives the response. A countervailing mechanism asserts that something is eroding a position; if margin, share or pricing power hold through the response, it was noise.

coverage and its limits

7 of 7 declared members resolve in the model. Effects are deliberately small: membership is cheap and the commitment that would matter is silicon.

Where it comes from

Generated from Ultra Ethernet Consortium — Ethernet reworked for AI scale-out, published by the parties who sell Ethernet against the party who sells InfiniBand. Members include AMD, Arista, Broadcom, Cisco, Meta, Microsoft and Oracle.

The instance that tests the mechanism rather than illustrating it. Every other coalition here forms because its members individually lack the power to set a standard — but Ethernet is the most widely deployed networking standard in existence, and Broadcom, Cisco and Arista are not weak in it.

So the reading has to be narrower and it is worth stating: the coalition is weak in the AI fabric specifically, not in networking generally. InfiniBand held the training-cluster interconnect and NVIDIA owns InfiniBand; UEC is the incumbent-elsewhere organising against an incumbent-here. That is a different shape from UALink, where the members have no position at all to defend.

Which makes it the useful control case. If coalition formation predicted failure on its own, UEC would be the counterexample. It is not — it predicts that the members lack power *in the contested layer*, and there UEC's members genuinely did.

Sources

Arguments about this mechanism

None yet. No thesis names this mechanism’s subject, which is a gap in the bank rather than a fact about the mechanism.

Arguments that run through it

These name a node on the path rather than the subject, so they pass through this mechanism without being about it. Shown separately and never graded as claims about it — hubs above the graph’s own 90th-percentile degree are excluded, or every argument touching NVIDIA would attach to every mechanism NVIDIA touches.