**Note** from Bead: Business Model And Strategy · [canonical source](https://redfish.acequia.io/guerin/.agents/fa3b0dc9-9dc9-479d-b40a-d6d5400ae313/2026-06-11/notes/02-china-minimum-requirements.md) · session 2026-06-11 · discussion: Talk: Business Model And Strategy
Working note crossing the [commercial-offerings survey](https://redfish.acequia.io/guerin/.agents/fa3b0dc9-9dc9-479d-b40a-d6d5400ae313/2026-06-11/notes/01-commercial-offerings-survey.md) with the [China market-entry report](https://redfish.acequia.io/guerin/.agents/fa3b0dc9-9dc9-479d-b40a-d6d5400ae313/2026-06-11/artifacts/china-market-entry-report.md). Not legal advice.
## The one-paragraph answer The legal minimum to sell software-only subscriptions to China is **nothing new**: host offshore, sell cross-border through a merchant of record (Paddle) or Stripe with Alipay/WeChat/UnionPay, keep PIPL hygiene (privacy notice, separate consent, stay under the 100k-individuals exemption), and screen customers against US export lists. ICP filing/license only attaches when you **host inside China**; charging subscribers from China-hosted infrastructure requires a commercial ICP (B25) license that effectively requires a Chinese partner or a pilot-zone WFOE. The two hard walls are product-specific: **China-territory geospatial data** (restricted to licensed entities, in-country map storage, GCJ-02, foreign firms can't hold map qualifications) and **government/SOE emergency-management procurement** (Document 79 / xinchuang ≈ closed to US vendors; the wedge is licensing the engine to a Chinese partner who fronts it as domestic).
## Per-offering implications | Offering | China fit | Minimum path | Hard constraints | |---|---|---|---| | **AgentScript** (JS ABM library) | Easiest | npm/CDN distribution; no ICP/service nexus; monetize support/training cross-border | China CDN mirror needs an ICP-filed partner; otherwise nearly unregulated | | **Simtable** (sand-table sim SaaS) | Medium | Tier 0 for private training/education buyers (zh manual already ships in v4.0); reseller for B2B fapiao | Latency across GFW hurts live ops use; live China incident geodata = restricted; gov/SOE EM market closed → partner-license the engine | | **realtime.earth** (real-time imagery/GIS) | Hardest | Offshore viewer rendering customer-own or global data only | Internet-map qualification unavailable to foreigners; map data must sit on China servers; GCJ-02 mandatory; citizen-imagery capture is surveying-adjacent → needs licensed Chinese map partner (Amap/Tianditu) | | **Acequia** (platform/account layer) | Structurally interesting | Sell the **pattern, not the platform**: the self-hosted "acequia-conformant stack" (bead `31bd5380` DIY path) is the natural China shape — partner runs it onshore, satisfying localization by design | acequia.io SaaS crosses the GFW (WebDAV+SW+WebRTC+TURN+tunnel broker); **tunnel/relay infrastructure resembles circumvention tooling — unlicensed VPN-like ops are illegal in China**; do not operate TURN/tunnel endpoints into China without counsel |
## Strategic observations 1. **The data-sovereignty pitch inverts in China.** At home, "agent location = agent loyalty" is an anti-cloud differentiator. In China the same architecture is the *compliance* answer: localization-by-design via a self-hosted acequia run by a Chinese licensee. The Azure-China / 21Vianet pattern is literally the acequia governance model — a local parciante operates the infrastructure under license. 2. **Sequence by friction:** AgentScript (now, ~free) → Simtable training/edu cross-border (Tier 0–1) → partner-licensed onshore acequia (Tier 2) → own license (Tier 3, only if volume proves out; pilot-zone B25 fit untested, 13 approvals total as of Feb 2025). 3. **Per-deal haircut to price in (cross-border B2B):** ~6% VAT withheld + 0–10% royalty WHT + reseller margin; MoR ~5% for B2C. 4. **2025 Network Data Regs** ask foreign PIPL-scope handlers to appoint a China representative — low enforcement risk at small scale, but it's the first thing a compliance-conscious Chinese enterprise customer will ask about. 5. **US-side controls are the silent step zero:** EM/public-safety buyers are disproportionately Entity-List/MEU-adjacent; screening must precede any deal regardless of tier.
## Open questions for Stephen - Which buyer do we actually want first: education/training (universities, fire academies) vs enterprise (industrial parks, insurers) vs the closed gov/SOE market via partner? - Is there an existing Chinese contact/partner candidate (the zh manual translation suggests prior demand)? - Appetite for the IP-leakage trade in the partner-license model?