high high likelihood
R-001: Financial opacity and runway cannot be validated publicly
Funding is public, but revenue, gross margin, burn, cash, debt, runway, backlog and AR aging are not public.
Diligence request: Request audited financial statements, management accounts, cash/debt schedule and three-year forecast before valuation work.
high high likelihood
R-003: Customer concentration and monetization not public
Public sources show product and logo signals but no verified top-customer list, revenue share, churn, or pipeline conversion.
Diligence request: Request top-15 customer schedule, cohort retention, signed contracts and reference calls.
high high likelihood
R-005: Hypercompetitive China foundation-model market
Baichuan competes with AI unicorns and large platform incumbents with distribution and compute advantages.
Diligence request: Run independent win/loss, benchmark and pricing comparison against Zhipu, Moonshot, 01.AI and platform models.
high medium likelihood
R-002: Valuation pressure in crowded LLM market
CB Insights valuation is $2.77B and Forbes references 20B yuan / $2.8B valuation basis; comparable Chinese AI unicorns and global benchmarks create repricing risk.
Diligence request: Reconcile valuation to current fully diluted cap table, preferences and updated operating metrics.
high medium likelihood
R-004: Compute, data and gross-margin dependency
LLM training/inference economics depend on compute, data rights and utilization; exact suppliers and costs are not public.
Diligence request: Review cloud/GPU contracts, model COGS, data licenses and utilization metrics.
high medium likelihood
R-006: China generative-AI regulatory compliance
Public generative AI services are subject to CAC-led Interim Measures; Baichuan has public model and healthcare-facing products.
Diligence request: Obtain filings, security assessments, content-safety policy, data governance and regulator correspondence.
high medium likelihood
R-007: Healthcare AI safety and liability exposure
百小医 and medical model positioning create risks around hallucination, medical advice, privacy and professional liability.
Diligence request: Review medical disclaimers, human-escalation workflows, safety evals, complaint logs and insurance coverage.
medium high likelihood
R-010: Product performance claims are largely self-reported
Benchmark, price/performance and download claims come mainly from company sources and technical reports.
Diligence request: Run independent benchmark, safety, latency, reliability and cost tests using current production APIs.