FRONTIER CN-1 / DISCOVERY ACTIVE · v0.1.0
China Discovery Universe
The first challenger to the US baseline. Twelve listed candidates, three private laboratories—and zero approved positions.
DISCOVERY UNIVERSE / NOT A PORTFOLIOChina may offer a different AGI expectation set: policy-driven deployment, domestic supply-chain substitution and large application markets—offset by ownership, access, governance, geopolitical and data-quality risk. This is a hypothesis to test, not a valuation conclusion.
The thesis weakens if investable securities do not capture the economics, governance discounts are not measurable, access costs erase the gap, or price expectations already discount the acceleration.
SECURITY IDENTITY / RESEARCH QUEUE
Candidate map v0.1
Inclusion means “research this security,” not “buy this security.” Every row begins with unresolved work.
Alibaba Group
9988 HK / BABA NYSE- research layer
- models · cloud
- Hypothesis
- Cloud infrastructure, model ecosystem and enterprise distribution may transmit AI adoption into revenue.
- blocking question
- How much owner cash is attributable to AI demand after capex, SBC and competitive pricing?
Tencent Holdings
0700 HK- research layer
- models · applications
- Hypothesis
- Models, cloud, advertising, games and social distribution create several AI monetization paths.
- blocking question
- Can incremental model and infrastructure spend earn above the existing business cost of capital?
Baidu
9888 HK / BIDU Nasdaq- research layer
- models · cloud
- Hypothesis
- Foundation models, AI cloud and autonomous-driving assets offer direct but capital-intensive exposure.
- blocking question
- Does AI revenue growth convert into durable segment cash economics rather than subsidized volume?
Xiaomi
1810 HK- research layer
- edge · physical · ai
- Hypothesis
- Large device distribution and electric-vehicle ambitions may turn models into edge and physical-AI adoption.
- blocking question
- Is AI an incremental margin and retention driver, or primarily a product-feature cost?
Lenovo Group
0992 HK- research layer
- edge · infrastructure
- Hypothesis
- Enterprise infrastructure, PCs and edge distribution may benefit from inference moving closer to users.
- blocking question
- Can AI infrastructure mix lift owner cash through a full hardware cycle?
Semiconductor Manufacturing International
0981 HK / 688981 SSE STAR- research layer
- semiconductor · foundry
- Hypothesis
- Domestic foundry capacity is strategically important to supply-chain substitution.
- blocking question
- What sustainable economics remain after technology constraints, capex intensity and policy support?
Hua Hong Semiconductor
1347 HK / 688347 SSE STAR- research layer
- semiconductor · foundry
- Hypothesis
- Specialty-node capacity can benefit from local industrial and edge-compute demand.
- blocking question
- Is utilization and return on new capacity resilient without extrapolating policy-led demand?
Cambricon Technologies
688256 SSE STAR- research layer
- compute
- Hypothesis
- Domestic accelerator design is a direct substitution candidate in constrained compute markets.
- blocking question
- Can the ecosystem, software stack and unit economics support durable adoption without assuming strategic importance equals shareholder return?
ZTE
0763 HK / 000063 SZSE- research layer
- networking · edge
- Hypothesis
- Telecom, networking and enterprise infrastructure sit on the path from compute to deployment.
- blocking question
- How exposed are cash flows to geopolitics, customer concentration and low-return infrastructure cycles?
Kingsoft Cloud
3896 HK / KC Nasdaq- research layer
- cloud
- Hypothesis
- Cloud demand provides a possible pure-play path into enterprise AI infrastructure.
- blocking question
- Can growth become self-financing after capex, pricing pressure and customer concentration?
GDS Holdings
9698 HK / GDS Nasdaq- research layer
- data · centers
- Hypothesis
- Data-center capacity and deliverable power are physical inputs to domestic model scaling.
- blocking question
- Do contracted demand, leverage and development capex produce acceptable equity cash returns?
VNET Group
VNET Nasdaq- research layer
- data · centers
- Hypothesis
- Carrier-neutral data centers may capture interconnection and compute-location demand.
- blocking question
- Can governance, financing and utilization risks be underwritten with sufficient confidence?
NON-LISTED COMPETITIVE MAP
Private laboratories / capability radar
These entities matter to the competitive map but do not provide direct exchange-listed ownership.
DeepSeek
Frontier-model capability and price/performance signal.
official source ↗ByteDance
Consumer distribution, models and high-volume inference demand.
official source ↗Huawei
Compute, networking, cloud, devices and domestic technology substitution.
official source ↗PRE-INVESTMENT CONTROL PLANE
Eight gates before a weight
Strategic importance is not shareholder return. Every candidate must pass all eight tests before portfolio construction.
Security identity and ownership
Verify legal claim, listing line, fungibility, custodian treatment and VIE/contractual exposure.
Market access
Confirm broker eligibility, Stock Connect rules, professional-investor limits and settlement.
Sanctions and export controls
Screen entity, affiliates, products and transaction path under applicable law.
Accounting and audit
Reconcile filings, auditors, related parties, consolidation and segment disclosures.
Governance and state intervention
Model control rights, capital allocation, policy constraints and minority-holder outcomes.
FX, tax and repatriation
Measure currency basis, withholding, convertibility and cash-movement frictions.
Liquidity and implementation
Estimate spread, impact, capacity, trading calendars, borrow and corporate-action risk.
Expectations and owner cash
Require point-in-time price expectations and cash economics after capex, dilution and financing.
RESEARCH PROTOCOL
From narrative to comparable evidence
- 01
Resolve the exact investable security and economic claim.
- 02
Ingest primary filings, policies and exchange notices with timestamps.
- 03
Separate AI revenue narrative from segment cash economics.
- 04
Normalize capex, dilution, leverage, FX and governance risk.
- 05
Build point-in-time expectations and falsification cases.
- 06
Only then design a matched US × China Alpha vs Alpha experiment.
PRIMARY-SOURCE MAP