Anthropic's business model is a product of the US cloud-native software market — a trinity of cloud delivery, standard agents, and PE distribution networks. Entering China means all three layers need to be completely redesigned: the deployment layer shifts to private, the product layer shifts to deep customization, and the distribution layer shifts to SOE trust chains. It's not impossible, but the transformation cost and timeline are severely underestimated.
The main competitor for a Chinese version of Anthropic is not DeepSeek or Zhipu, but Yonyou and Kingdee. The data is in the ERP, the trust is in the ERP, and the decision-making budget is in the ERP. Before entering large enterprise clients, LLM companies must first answer a prerequisite question: embed into the ERP, or replace the ERP from the outside? The two paths have completely different margins, sales motions, and capital structures — there is no middle ground.
To assess the portability of the Anthropic model, we must first establish the software industry baselines of three markets — not market size, but structure: business models, deployment preferences, and revenue composition. The gap between these three markets is larger than most people realize.
Europe is the closest reference to China among the three: data sovereignty constraints (GDPR ≈ Data Security Law + PIPL), higher government procurement share, and stronger On-Prem preference — these three points share structural similarities with China. But China adds three unique layers on top: LLM filing requirements, Xinchuang domestic substitution mandates, and the political vetting chain in SOE procurement. European adaptation experience is useful as a reference, but cannot be directly copied.
New deployments default to cloud; AWS/Azure hosting is the standard path. Private deployment is the exception, seen only in government classified or special sectors (defense/nuclear)
Mittelstand ~40% still prefer On-Prem; GDPR requires data to stay within specific jurisdictions, driving "sovereign cloud" (Deutsche Telekom/OVHcloud) demand
Private deployment is the default, not the exception. Central SOEs/state-owned banks/critical infrastructure data cannot leave local server rooms — this is a hard compliance requirement, not a negotiating chip. 64% of large enterprise AI procurement requires private deployment
→ Anthropic's cloud API delivery model fails directly in China's large enterprise market.It must provide a privately deployable version that can run independently in customer data centers, which means: GPU procurement/leasing costs shift from cloud to vendor, fundamentally changing the margin structure.
"Fit to standard" culture: enterprises are willing to adapt internal processes for standard products. Salesforce and Workday's success is built on the philosophy of "configuration, not development"
German SMEs in particular lean toward customization: the "Made in Germany" spirit extends to software, with lower willingness for process standardization; SAP's rise itself served this demand
SOE/central SOE projects can reach 80%+.Historically, due to rampant ERP piracy, Yonyou/Kingdee built competitive moats through extensive customization services; enterprise clients thus became accustomed to the mindset of "software is a custom product"
Subscription-first:Salesforce seat fee $25–300/month, Anthropic Claude Enterprise API by token consumption + fixed seat fee, revenue is predictable and scalable
Light implementation:GSIs charge implementation fees, product vendors don't participate in implementation, margin structure is clear
Upgrade path:Automatic version updates, invisible to customers, marginal cost approaches zero
Perpetual license dominant:Enterprises are accustomed to one-time buyouts; in Yonyou's FY2024 revenue of 9.15B RMB, subscriptions accounted for only ~20%, the rest being license + implementation fees
Implementation revenue is a profit center:Software license : implementation fee = 1:3 to 1:5 (vs. ~1:1 in the US), implementation teams are the core competitive advantage
High upgrade resistance:Extensive customization code means version upgrades require re-implementation, customers tend not to upgrade — generating massive "implementation debt"
→ Pure subscription / pure API consumption pricing logic requires a hybrid model in China's large enterprise market: subscription base + implementation services + custom development, which means building implementation delivery capabilities
Anthropic's PE JV model works in the US because Blackstone/KKR etc. hold 2,000+ portfolio companies; a JV partnership equals opening procurement channels to these companies simultaneously. This structure does not exist in China.
Blackstone portfolio companies 2,700+, one JV agreement is equivalent to issuing entry tickets to 2,700 companies. GPM unaffected, distribution marginal cost extremely low
Capgemini/Sopra Steria are the main distribution channels; Europe lacks equivalent PE networks, but GSI penetration is high, covering large enterprises through GSI partnerships is relatively feasible
No independent PE network; the core of distribution is the SOE trust chain: ICBC/China Mobile/State Grid procurement → other central SOEs reference → leading private enterprises follow. Building the trust chain takes 2–4 years
CTO/CIO-led, technical POC is standard process; Europe adds Data Protection Officers (DPO) and legal review, but the overall path is clear. Dimon's endorsement = JPMorgan legal compliance sign-off = industry entry ticket
Typical large SOE procurement path: Technical department research → IT committee review → Information security compliance audit (MLPS, data security) → Procurement committee → (Party committee) study → Leadership approval. Every step can become a bottleneck
→ China's version of "Dimon's endorsement" = ICBC president's public endorsement; but it's far harder to obtain than in the US, and SOEs typically don't simultaneously procure competing products — once secured, the lock-in effect is stronger
Cloud-native is the default path; Anthropic agents delivered via API/cloud subscription, no local deployment needed, enterprises can self-serve onboarding
Germany/France on the lower side (~18–22%), Nordics higher (~35%); GDPR drives sovereign cloud demand rather than hindering SaaS overall
Budget concentrated in large enterprises, which generally require private deployment; SaaS growth mainly from mid-sized private enterprises' non-core applications
| Metric | 🇺🇸 US (Current) | 🇨🇳 China (Adapted Estimate) |
|---|---|---|
| Delivery Model | Cloud API / Claude.ai Teams | Private deployment + Cloud API (Hybrid) |
| Gross Margin (GPM) | ~70–75% | ~30–40% (Private deployment dilution) |
| Typical Enterprise Contract Size | $500K–5M / year | ¥3M–15M / year (~$0.4–2M) |
| Sales Cycle | 3–6 months | 12–24 months |
| Implementation Cost (Vendor-Borne) | Low (GSI bears it, vendor doesn't participate) | High (Must build implementation team or deeply partner with SI) |
| Pricing Basis | Token consumption + Seat subscription | Annual subscription + Custom dev fee + Implementation fee |
| Standard Configuration Ratio | ~80% (Low customization) | ~30% (High customization) |
| Time to Scale to 100 Large Enterprises | ~18–24 months | ~4–6 years |
Anthropic's enterprise AI business model consists of three interdependent layers. Understanding why each layer works in the US is essential to precisely determining which layer will fail first in China.
Enterprises can directly access Claude via the Anthropic API, no local deployment needed. Cloud hosting means Anthropic's compute costs are amortized across Google/Amazon's infrastructure, GPM ~70%+. Token billing enables "pay for what you use," lowering the procurement barrier
Cloud API is ineffective for the 64% of large enterprises requiring private deployment. Even accepting a hybrid model, LLM service filing requires domestic services to use registered models. Foreign models face compliance barriers providing services within China; Anthropic itself cannot directly enter core scenarios in API form
US financial industry report formats (SEC 10-K/10-Q), compliance processes (SOX/Dodd-Frank) are highly standardized. 10 financial agents can cover 80% of JPMorgan's common scenarios without customization. GSIs handle the remaining 20% adaptation
Chinese financial industry report formats are fragmented (A-share/HKEX/CBRC/CSRC/PBOC formats are mutually inconsistent), OA approval workflows are highly personalized, supply chain data standards lack unification. Standard agents' "out-of-the-box" capability fails before entering the meeting room — every enterprise will say "our situation is somewhat unique"
A JV agreement with Blackstone etc. is equivalent to simultaneously opening procurement channels to 2,700+ portfolio companies. Dimon's public endorsement = JPMorgan legal compliance has reviewed = other CTOs in the financial industry can reference this precedent to reduce their own decision risk
Chinese PE/VC don't hold operating enterprises' procurement rights, can't "bundle" channels. Trust signal sources are different: SOE CEO statements > private CEO endorsements. Building an SOE trust chain requires not JV equity structures, but real deployment cases + government procurement records — this takes 2–4 years to accumulate
→ Layer 3 is the hardest to transform among the three layers. Layer 1 and Layer 2 can be solved through technical investment; Layer 3 requires time and political capital, neither of which can be bought by spending more money.
Anthropic enters the Chinese market directly with its current cloud API model + pre-built agents + PE JV structure.
Cloud API cannot enter private deployment scenarios; overseas model filing barriers; no cross-border data transfer
US financial agent formats don't match Chinese reporting standards; no native Chinese workflow support
PE JV has no equivalent structure in China; no SOE trust chain; compliance credit is zero
Conclusion: Can only serve a small number of foreign enterprise China branches and export-oriented leading private enterprises; cannot enter government/finance/SOE core scenarios, which account for 60%+ of China's AI spending
Anthropic or an equivalent international AI company establishes an independent legal entity in China, partnering with domestic collaborators to: develop a private deployment version, rewrite Chinese agents, and build an SOE JV structure. This corresponds to Google/Microsoft's past China strategies.
Timeline: Earliest entry into non-core central SOE scenarios in 3 years, core finance/government scenarios in 5–7 years. Capital requirement: ~1–3 billion RMB build-out investment
Chinese AI vendors replicate Anthropic's three-layer structure, but fill each layer with domestic resources: domestic LLMs + Chinese enterprise agents + SOE/cloud vendor distribution networks. This is the most likely path to success.
DeepSeek / Kimi / Qwen / Wenxin · Filed · Support private deployment · Xinchuang adaptation feasible
Rewrite A-share report agents / SOE OA agents / Supply chain agents; embed Yonyou/Kingdee data interfaces
Central SOE strategic investment + Alibaba Cloud/Huawei Cloud pipeline + Yonyou/Kingdee ISV channels; saves 2–3 years of trust building vs. foreign vendors
Conclusion: No single company currently possesses all three layers simultaneously. The "fastest winner" of the Anthropic model in China may be a new JV entity, rather than any existing single vendor
The following eight changes are mandatory — missing any one reduces the score. The first four are at the technical and product level, the latter four are at the business and compliance level — the latter are actually harder, because you can't trade engineering capability for time.
Must simultaneously provide three deployment forms: ① Public cloud version (mid-sized private enterprises, non-core applications) ② Private deployment version (large enterprises/finance/SOEs) ③ Xinchuang version (SOEs/government, running on FeiTeng/Kunpeng CPU + Kylin OS). The three SKUs are not simple repackaging of the same codebase — the Xinchuang version requires adaptation for non-x86 architectures, with engineering effort approximately 30–50% of building a new product
Shift from pure subscription/consumption model to three-part pricing: ① Annual platform subscription base (recurring revenue source) ② Custom development fee (priced by module, not by person-day) ③ Implementation service fee. Key: custom development fees must be productized (with fixed SKUs), otherwise it devolves into a zero-margin project company rather than a scalable AI platform
Not translating US agents, but native rewrites: A-share annual report analysis agent (connecting to Wind data), SOE OA approval agent (connecting to DingTalk/WeCom workflows), supply chain agent (connecting to Yonyou/SAP data fields), bank anti-fraud/intelligent customer service agent (connecting to PBOC data standards). Each agent's underlying data format is based on Chinese standards, not translated versions of US standards
Provide a low-code customization framework rather than fixed templates: client IT or implementation partners can drag-and-drop assemble within the framework, without vendor engineers needing to participate in every project. This is the key to shifting "customization debt" from vendor to client/channel. Reference Salesforce Flow / ServiceNow's no-code configuration layer logic
Four compliance certifications are all indispensable: ① LLM service filing (CAC, prerequisite for providing generative AI services) ② MLPS Level 3 certification (minimum threshold for entering finance/government scenarios) ③ PIPL/DSL compliant data processing agreement ④ Xinchuang compatibility certification (joint testing with specific hardware vendors). Filing cycle ~3–6 months, MLPS certification ~6–12 months, must start early
The Chinese version of "PE JV" is a triangular partnership structure: ① Central SOE industry fund (providing compliance credit + SOE client procurement endorsement, taking minority stake) ② Top cloud vendor (Huawei Cloud/Alibaba Cloud, providing compute infrastructure + cloud marketplace distribution) ③ ERP vendor strategic partnership (Yonyou/Kingdee, data interfaces + existing customer reach). Missing any side of the triangle significantly slows entry into core scenarios
The only viable trust chain building sequence:Government scenario reference customers (regardless of how small the revenue, first create publicly promotable cases) → Policy banks / State-owned major banks → Central SOEs → Leading city commercial banks / insurance → Leading private enterprises → Mid-sized enterprises. There is no shortcut that bypasses "government first." Anthropic's Dimon path works in the US because US private financial institutions are highly independent; China needs government nodes as trust anchors
Two things from Europe are worth directly borrowing: ① Sovereign cloud architecture (GDPR-compliant data localization implementation can directly correspond to DSL/PIPL requirements, reducing redundant development) ② German SI ecosystem model (Germany handles high customization demand by strengthening SI partnerships rather than product standardization, which is closer to China's demand structure). What's not worth borrowing is Europe's AI Act framework — the EU's risk-tiering logic is fundamentally different from China's filing system logic, and cannot be directly mapped
The CAC may require AI agents to be filed separately (rather than just the underlying model); compliance costs multiply, and requirements may change with policy cycles, creating retroactive adaptation costs
If Xinchuang requirements extend from government/SOEs to the entire financial industry or critical infrastructure, the hardware ecosystem available for AI deployment will significantly narrow, and Nvidia GPU-based private deployment solutions may become invalid
Each private deployment project requires the customer or vendor to procure GPU clusters (millions to tens of millions of RMB scale). As the number of private deployment projects increases, GPU procurement/maintenance costs become an explicit liability, while the scale effects of the cloud API model cannot be transferred
Yonyou YonGPT and Kingdee AI are growing AI capabilities from within the ERP. Despite Yonyou's FY2024 loss of 2.06 billion RMB, it controls data access for 7 million enterprises — external LLM vendors need to prove "why not just use the ERP's built-in AI"
Highly customized projects result in each client being an independent code branch; version upgrades require re-implementation. When custom clients exceed 50, maintenance costs may exceed new business revenue, forming a "growth trap"
Hybrid talent who understand both LLMs and enterprise ERP/OA systems, and can communicate with SOE clients, is extremely scarce. Unlike the ERP era, AI implementation requires greater technical depth, but the existing SI ecosystem has not yet developed this talent pool
The Anthropic model works because it highly aligns with the structure of the US software market: cloud delivery corresponds to the cloud-native deployment default, standard agents correspond to the 25% customization rate, and PE JV corresponds to the private PE enterprise network. There is no friction between the three layers — each layer paves the way for the next.
The six structural differences in China's software market cause this alignment to disappear entirely. It's not that one layer doesn't match — it's that all three layers don't match, and the reasons for each layer's mismatch are different, so they cannot be solved with the same approach.
This doesn't mean there's no market in China — the geographic moat formed by data sovereignty actually creates a domain for domestic AI service providers that US competitors can never encroach upon. But entering this domain requires rebuilding the three-layer architecture using the logic of Chinese industry, not translating Anthropic's three layers.
European experience provides partial reference — sovereign cloud architecture, SI-led customization handling, government-procurement-first trust chains — but China's regulatory density and SOE-dominated landscape mean European cases can only be referenced, not copied.
One final judgment: China does not lack model capabilities (Layer 1). What it lacks is a combination that can simultaneously do Layer 2 (Chinese agents) and Layer 3 (distribution trust) well. Whoever can put these two layers together will be China's Anthropic — but it won't be called Anthropic, and it won't be led by an American company.
First published 2026-05-09