EN / 中文

Chapter 13 / 16

Core, Watch, and Warning Pools

Separate Core, Watch, Crowded Winners, Wait for Pullback, Low-Consensus Bottlenecks, and Narrative Avoidance pools.

Current Qualitative Consensus Map

As of 2026-06-20, based on recent market and industry signals:

ThemeBottleneckConsensus / priced-inDiscovery gapNotes
Nvidia GPU / full-stack AI compute550Real bottleneck, but fully consensus. Use as demand signal, not hidden opportunity.
HBM leaders: SK hynix, Micron, Samsung54-50-1Very real, but now widely recognized. Still cyclical upside possible, less undiscovered.
Nearline HDD: Seagate, Western Digital43-41AI storage demand is being discovered; not hidden anymore, but less crowded than HBM.
CXL / memory pooling: Astera, Rambus, Montage Technology42-31-2Better non-consensus layer: key if memory shifts from fixed server DRAM to pooled memory.
DDR5 interface / retimer: Montage Technology, Rambus, Astera42-31-2More attractive than generic memory modules because it sits in the control/interconnect layer.
Enterprise SSD / NAND controllers: Silicon Motion, SanDisk, Longsys, Biwin Storage3-430-1Good cycle trade, but more exposed to commodity pricing and inventory swings.
AI host CPU attach: AMD, Arm, Nvidia Grace/Vera43-40-1Becoming visible; still worth tracking because GPU rack architectures force CPU attach.
Intel CPU / foundry / packaging turnaround3-430-1Potentially large, but execution risk is high; separate turnaround thesis from AI bottleneck thesis.
A-sharedomesticCPU: Hygon Information Technology, Loongson Technology3-43-40-1More policy/import-substitution than pure AI bottleneck; valuation often already anticipates this.
AI server OEM/ODM: Inspur Electronic Information, Sugon, Foxconn Industrial Internet33-4-1-0Demand is obvious, margins may be competed away; watch order quality and profit capture.

Automation Label

Each company should receive one of these labels:

  • crowded_winner: real winner, but already heavily priced.
  • watch_for_pullback: good business, weak margin of safety at current consensus.
  • underdiscovered_bottleneck: best target for deep research.
  • cycle_beta: mostly memory / hardware cycle exposure.
  • execution_option: upside depends on company execution, not only industry bottleneck.
  • avoid_narrative_only: AI story exists, but bottleneck capture is unclear.

Initial labels:

CompanyMarketLabelRationale
NvidiaUScrowded_winnerStrongest AI infrastructure position, but fully consensus.
MicronUSwatch_for_pullbackHBM/DRAM/NAND bottleneck is real, but memory boom is now widely recognized.
SeagateUSwatch_for_pullbackNearline HDD scarcity is increasingly recognized; still less crowded than HBM.
Western DigitalUSwatch_for_pullbackSimilar HDD/storage thesis; monitor pricing discipline and supply commitments.
Astera LabsUSunderdiscovered_bottleneckCXL/retimer/connectivity layer is strategically important, but valuation may already embed some scarcity premium.
RambusUSunderdiscovered_bottleneckMemory-interface IP and chips may benefit from DDR5/CXL without taking full commodity memory risk.
AMDUSwatch_for_pullbackCPU attach and AI accelerator upside are real, but increasingly consensus.
Arm HoldingsUSwatch_for_pullbackServer Arm attach is real, but valuation often prices a broad architecture win.
IntelUSexecution_optionFoundry/packaging/CPU recovery has upside, but thesis depends on execution.
Montage TechnologyA-shareunderdiscovered_bottleneckDDR5 interface, PCIe retimer, CXL MXC; better positioned in memory interconnect than generic storage names.
Hygon Information TechnologyA-sharewatch_for_pullbackDomestic CPU leader; strong policy/domestic substitution narrative already partly priced.
SugonA-sharecycle_betaAI server/system demand beneficiary; profit capture needs verification.
Inspur Electronic InformationA-sharecycle_betaHigh AI server relevance, but more competitive hardware margin profile.
LongsysA-sharecycle_betaStorage cycle and module upside, but more exposed to commodity swings.
Biwin StorageA-sharecycle_betaSimilar storage cycle beta; needs evidence of durable enterprise/AI mix.

Primary automation sort:

research_priority = monopoly_potential_score + discovery_gap + evidence_velocity - valuation_heat

Automation should surface:

  • New high-priority company mentions.
  • Companies with rising evidence velocity.
  • New bottleneck terms that appear across multiple sources.
  • Disagreements between investor narrative and operational evidence.
  • Companies where consensus score rises faster than evidence velocity, which may mean the trade is becoming crowded.

Observation Pool, Not Core Pool

  • SIVE / Sivers Semiconductors: high-signal photonics bottleneck candidate, supported by Jabil and GF collaborations, but social heat and valuation expectations are high. Need independent proof of share, gross margin, capacity rights, and product revenue ramp.
  • AAOI / Applied Optoelectronics: repeatedly appears in Serenity-derived photonics discussions, but needs a narrow monopoly layer beyond "optical networking demand."
  • POET / Celestial AI / Ayar Labs ecosystem names: strategically important to CPO / optical compute narratives, but public equity capture and supplier economics differ by company.
  • PENG / Penguin Solutions: useful CXL product evidence; core inclusion requires proof that Penguin captures margin versus being a systems integrator.
  • MRVL / Marvell: strong AI connectivity and custom silicon narrative; likely more consensus now. Treat as platform winner/watch_for_pullback unless a specific unavoidable layer is isolated.
  • NBIS / neocloud and powered data-center names: contracted power is valuable, but equity value depends on active MW, utilization, customer credit, and financing terms.
  • Leader Harmonious Drive Systems: humanoid/physical AI component thesis is plausible, but this week's move looks social-liquidity driven; separate from AI data-center radar.

Warning Pool

  • Any Serenity-mentioned A/H stock that gaps before primary evidence appears. 21jingji's Leader Harmonious Drive Systems report is a case study in social heat causing immediate price action.
  • SIVE if the thesis is reduced to "Serenity likes it" or if market cap prices in 20-30x revenue growth before volume revenue is visible.
  • Generic storage names after broker target hikes and media stories about sold-out HDDs. Real shortage, but discovery_gap is shrinking.
  • AI data-center power names with only headline GW/MW claims and no interconnection, PPA, permitting, or active-load evidence.
  • CXL names where the product is technically interesting but no design win, attach-rate, or ASP/margin evidence exists.