China manufactures roughly 30 percent of global goods by value and a considerably higher share of electronics. For a deep-tech hardware company — one building novel sensors, medical devices, embedded systems, or AI-enabled physical products — there is often no realistic alternative to engaging some part of that supply chain. The question is not whether to do so, but how to do so in a way that does not hand the core of your business to the first competent manufacturer who inspects your Gerber files.

The IP risks in contract manufacturing are real, but they are frequently mischaracterised. The popular narrative — that Chinese factories will inevitably clone your product — is too blunt to be useful. The actual risks are structural and mostly preventable with the right preparation. Understanding them precisely is the first step.

Design leakage at the quoting stage is the most common and least appreciated risk. When you send a full bill of materials, schematics, and firmware binaries to five factories for competitive quotes, you have already distributed your IP before any contract is signed. Many founders treat quoting as a pre-commercial interaction governed by goodwill. It is not. A complete set of production documents handed to the wrong contact is all a well-resourced actor needs.

Tooling ownership ambiguity is the second structural risk. Custom injection moulds, stamping dies, and fixture jigs are expensive to fabricate and encode significant design information. Unless contracts explicitly state that tooling is owned by the buyer and held on consignment at the factory, the legal default in many jurisdictions — including China — treats tooling paid for by the manufacturer as manufacturer property. That creates a lock-in that makes switching factories far more costly than it appears at the outset.

Grey-market overruns are a third category that sits between contractual breach and IP theft. A factory running a production line at contracted volume incurs most of its cost in setup and tooling amortisation. Marginal units cost very little. The incentive to run additional units beyond the contracted quantity, and route them to grey-market channels, exists whenever the contract does not specify production audit rights and does not impose credible penalties.

Against these risks, the most effective protections are structural rather than purely contractual.

NNN agreements — non-disclosure, non-use, and non-circumvention — are the appropriate instrument for pre-commercial engagement in China, not the Western NDA. An NDA is oriented toward confidentiality in a disclosure context. An NNN agreement additionally prohibits the recipient from using your information to compete and from circumventing your business relationship to deal directly with your customers or suppliers. The distinction matters. Register the agreement in Chinese and have it governed by PRC law in a PRC jurisdiction — an English-law contract is of limited practical utility if enforcement ever becomes necessary in a Chinese court.

Design partitioning reduces the surface area any single factory can exploit. The principle is straightforward: split the product so that no single manufacturing partner holds a complete, production-ready design. A common approach is to have the most IP-sensitive subassembly — the custom ASIC, the proprietary sensor array, or the firmware-loaded module — manufactured in a separate jurisdiction or at a separate facility, and integrated only at final assembly. The final assembler sees a black-box component; the subassembly manufacturer sees no finished product context. Neither holds enough to replicate the whole.

Split supply chain management complements partitioning. Rather than having the factory manage all component procurement, the buyer retains direct relationships with suppliers of the most critical and differentiated parts — particularly any component that carries firmware, contains a unique algorithm in silicon, or whose specification itself constitutes trade secret. When the factory procures generic commodity components and the buyer supplies the critical ones, the factory's ability to replicate the product independently is materially reduced. This is one area where working with a trusted compliance and sourcing gateway — rather than managing the supply chain entirely through a local intermediary with mixed allegiances — provides a structural advantage. A gateway with audited supplier relationships and a vested interest in the buyer's long-term success has different incentives than a local agent whose income depends on factory commissions. Asaptic covers the principles of this kind of structured engagement in the deep-tech sourcing overview.

Registering IP in China is underutilised by foreign founders and genuinely effective. China's patent and trademark systems are functioning courts with real enforcement track records for registered rights. A patent registered in China by a foreign entity is enforceable in China against a Chinese infringer. Trademark squatting — where a third party registers your brand name before you do — is a known hazard that is entirely avoidable by filing early. The cost of registration is low relative to the cost of enforcement after the fact. The cost of not registering is incalculable if the product succeeds.

Controlled BOM and firmware practices close the loop at the technical level. Firmware shipped to a factory should be locked to a specific hardware identifier where possible — so that a binary pulled from a device during production cannot simply be reflashed onto a clone board. Critical algorithm parameters should be provisioned post-assembly through a secure channel, not baked into the shipped firmware image. Production BOM documentation shared with the factory should reference internal part numbers, not open-market specifications, wherever possible — slowing the factory's ability to independently source equivalents.

None of these measures are absolute. A sufficiently motivated and resourced competitor can eventually reverse-engineer any product at the component level, and contract enforcement across jurisdictions remains genuinely difficult. The realistic goal of an IP protection strategy is not imperviousness but friction — raising the cost and time required for imitation high enough that your product's market window, your rate of iteration, and your customer relationships provide the real competitive buffer. Most IP leakage events are opportunistic, not targeted. Structural protections remove the opportunity.

The one protection that matters most is also the one most commonly neglected at the early stage: working with counterparts who have a demonstrated, audited track record of compliance — and who have more to lose from a breach than they would gain. Due diligence on the integrity of your manufacturing relationships is not a legal formality. It is the foundation on which every other protection rests. See the deep-tech sourcing principles for how a structured gateway model addresses this at the relationship level.

This article is educational and general in nature. It does not constitute legal advice. IP strategy for a specific product and jurisdiction should be developed with qualified legal counsel.

在中国进行合同制造时，知识产权面临的真实风险往往被误读。常见的叙事——"中国工厂必然会克隆你的产品"——过于笼统，实际上无助于制定有效的应对策略。真正的风险是结构性的，且大多可通过充分准备加以预防。

报价阶段的设计泄露是最常见也最易被忽视的风险。将完整的物料清单、原理图和固件发送给多家工厂进行报价，意味着在任何合同签署之前，你的核心设计已被广泛分发。模具所有权归属不明是第二大结构性风险——除非合同明确约定模具由买方所有并寄存于工厂，否则在多个司法管辖区（包括中国），由制造商出资制作的模具在法律上归制造商所有，这将大幅提高更换工厂的成本。灰市超产则游走于合同违约与知识产权侵害之间：当合同未规定生产审计权且缺乏有效惩罚机制时，工厂超额生产并通过灰市渠道销售的动机始终存在。

针对上述风险，最有效的保护措施是结构性的，而非单纯依赖合同条款。在中国应使用 NNN 协议（保密、禁止使用、禁止规避），而非西式 NDA；协议应以中文签署并适用中国法律管辖。设计分拆可将单一工厂能够利用的知识产权范围降至最低——将最敏感的子组件在独立设施生产，最终装配厂只接触黑盒零部件。关键零部件的分供管理可进一步削弱工厂独立复制产品的能力。在中国本地注册专利和商标是被众多外资创始人忽视却切实有效的保护手段。受控的 BOM 和固件分发实践——如将固件绑定至特定硬件标识、在装配后通过安全渠道进行关键参数配置——则从技术层面封堵漏洞。

没有任何措施能做到绝对防护。IP 保护策略的现实目标是提高仿制的成本与时间门槛，而非追求无懈可击。选择具有合规记录、且在关系层面拥有充分动机维护你利益的合作方，是其他一切保护措施的基础。

在中國進行合同製造時，知識產權面臨的真實風險往往被誤讀。常見的敘事——「中國工廠必然會克隆你的產品」——過於籠統，實際上無助於制定有效的應對策略。真正的風險是結構性的，且大多可透過充分準備加以預防。

報價階段的設計洩露是最常見也最易被忽視的風險。將完整的物料清單、原理圖和韌體發送給多家工廠進行報價，意味著在任何合約簽署之前，核心設計已被廣泛分發。模具所有權歸屬不明是第二大結構性風險——除非合約明確約定模具由買方所有並寄存於工廠，否則在多個司法管轄區（包括中國），由製造商出資製作的模具在法律上歸製造商所有，這將大幅提高更換工廠的成本。灰市超產則游走於合約違約與知識產權侵害之間：當合約未規定生產審計權且缺乏有效懲罰機制時，工廠超額生產並透過灰市渠道銷售的動機始終存在。

針對上述風險，最有效的保護措施是結構性的，而非單純依賴合約條款。在中國應使用 NNN 協議（保密、禁止使用、禁止規避），而非西式 NDA；協議應以中文簽署並適用中國法律管轄。設計分拆可將單一工廠能夠利用的知識產權範圍降至最低——將最敏感的子組件在獨立設施生產，最終裝配廠只接觸黑盒零部件。關鍵零部件的分供管理可進一步削弱工廠獨立複製產品的能力。在中國本地註冊專利和商標是被衆多外資創辦人忽視卻切實有效的保護手段。受控的 BOM 和韌體分發實踐——如將韌體綁定至特定硬件識別碼、在裝配後透過安全渠道進行關鍵參數配置——則從技術層面封堵漏洞。

沒有任何措施能做到絕對防護。IP 保護策略的現實目標是提高仿製的成本與時間門檻，而非追求無懈可擊。選擇具有合規記錄、且在關係層面擁有充分動機維護你利益的合作方，是其他一切保護措施的基礎。