（来源：CHINA DAILY 2026-07-27）
Illustration generated by AI
As the era of artificial general intelligence approaches, global value chains in high-technology services are being reconfigured, reshaping how tasks are organized across borders, where comparative advantages exist and who captures the value created.
Although global value chains are still commonly understood in terms of physical products assembled from components made in different countries, this conventional view increasingly fails to capture an important part of the global economy, which now moves through fiber-optic cables, cloud platforms and multinational corporate networks rather than in containers.
According to the World Trade Organization, global exports of commercial services reached $9.56 trillion in 2025, an increase of 8 percent over the previous year. Exports of digitally delivered services rose by 10 percent to $5.26 trillion. Computer services exports reached $1.22 trillion, more than twice their 2019 level. These flows are becoming increasingly central to international trade and globalization.
The globalization of high-tech services is best understood not simply as trade in finished services, but also as the global organization of problem-solving. Firms divide complex production and innovation processes into tasks, group those tasks into business functions, distribute them across countries and govern the connections among them.
The intellectual starting point for understanding this transformation is the "trade in tasks" framework developed by Gene Grossman and Esteban Rossi-Hansberg in their 2008 article "Trading Tasks: A Simple Theory of Offshoring". Their key insight was that advances in information and communication technologies allow firms to fragment production more finely.
This matters especially for services. Traditional economic thinking tended to treat services as non-tradeable because their production and consumption often occurred in the same place and at the same time. Yet many modern services activities — such as software testing, data analytics, chip design and remote equipment diagnostics — do not require such proximity.
A job is therefore better understood as a bundle of tasks. Once a task becomes digitally deliverable and organizationally separable, where it is performed becomes an international sourcing decision.
However, firms seldom relocate tasks in isolation. Research by the Organization for Economic Co-operation and Development shows that they often relocate bundles of related tasks organized as business functions, such as IT support, R&D, engineering, marketing and back-office services.
Fragmentation also creates interfaces, raises monitoring costs and adds to the risks of knowledge loss. It occurs only when the gains from specialization exceed the costs of coordination and governance. The technical trade-ability of a task therefore does not automatically produce organizational fragmentation.
The same logic helps us understand the architecture of high-tech service value chains, which resemble multilayered stacks. At the bottom are enabling inputs and infrastructure such as semiconductors, servers, data centers, energy and telecommunications networks. Above them are cloud platforms, databases, software tools and foundation models. Further up are coding, model training, chip design and data analysis. At the top are system integration, domain solutions, intellectual property and customer relationships.
Tasks can be distributed across many countries, while control over platforms, data, standards, intellectual property and customers remains concentrated in a small number of leading firms and technology hubs.
Yet the economic importance of these service layers is not fully captured by conventional trade statistics. According to the trade-in-value-added data for 2018 published by the WTO and the World Bank, services accounted for 50 percent of world trade in value-added terms. Services value added represented, on average, 31 percent of manufacturing exports in OECD economies and 29 percent in non-OECD economies. Services are increasingly the operating system of manufacturing value chains.
What, then, determines where these activities are located and how they are governed?
The first wave of services offshoring was largely driven by wage differences, whereas the location of high-tech service tasks increasingly depends on scarce skills, reliable digital infrastructure, data and computing power, regulatory compatibility, intellectual-property protection and dense innovation ecosystems. The underlying logic is therefore shifting from cost arbitrage toward capability arbitrage.
Evidence from business surveys illustrates this shift. In a 2023 Bain survey of more than 500 senior executives, 60 percent reported that their companies planned to increase engineering and research and development outsourcing over the following three years. Seventy-three percent identified industry or technology expertise as a leading factor in choosing a supplier, compared with 59 percent citing cost.
Cost advantages increasingly need to be combined with technological capabilities, organizational trust and responsibility for outcomes. The upgrading path appears to run from executing standardized tasks, to managing a business function, and then to owning product modules. Yet participation in global value chains does not guarantee value capture. What matters is who defines the architecture, who owns the intellectual property and who controls customer relationships.
Trade in tasks does not necessarily take the form of arm's-length outsourcing. A multinational can move work to a foreign affiliate, contract with a services provider, hire through an online platform, or purchase cloud and AI services. Routine and measurable tasks are easier to specify and contract out. Complex and interdependent activities are more likely to remain within the firm or be conducted through stable partnerships. Work involving strategic data, cybersecurity or core intellectual property may be technically tradeable, but organizationally non-tradeable.
Geopolitical considerations and circumstances increasingly shape these choices. Data-localization rules, privacy regulations, export controls, cybersecurity requirements and differences in AI governance affect the movement of service tasks. In goods trade, the border is visible at customs; in digital task trade, it may be embedded in a data architecture, a cloud contract or a technical standard.
High-tech service value chains may therefore become more geographically distributed but institutionally segmented, as firms selectively reconfigure them around trust, security and regulatory compatibility.
As artificial general intelligence draws closer, this transformation may accelerate. AI creates a fundamental paradox for trade in tasks: it can both facilitate cross-border work and replace it.
On the one hand, AI lowers the costs of trading tasks across borders. It not only improves communication and standardization through AI-powered translation and coding tools, but also helps firms monitor quality and coordinate geographically dispersed operations. On the other hand, AI automates many of the tasks that were easiest to offshore, including routine coding, translation, data entry, document review and standardized customer support.
Preliminary evidence from some cross-border labor platforms points to such a reorganization. Following the emergence of generative AI, outsourcing has declined in highly exposed skill areas relative to less-exposed areas, but the remaining contracts have become more complex and valuable, and work has become more concentrated.
The likely result is a hollowing out of routine digital tasks and a rising premium on human judgment, interaction, domain knowledge, system integration and responsibility. In other words, AI is changing comparative advantages at the level of the task. It substitutes for some cognitive activities while complementing others within the same occupation.
The policy implications follow directly from this analysis. Countries should not build a high-tech services strategy around low wages alone. Policy should instead support movement from tasks to functions and from functions to assets. Performing isolated tasks generates income; managing an end-to-end engineering function fosters learning; and owning software, patents, standards, data products or customer relationships creates scalable assets and greater bargaining power.
Services and industrial strategies must also be designed in tandem, as the competitiveness of electric vehicles, semiconductors, medical devices and industrial equipment increasingly depends on software, engineering, data analytics, cloud systems and after-sales services. Realizing these complementary aspects requires access to cross-border data flows and global cloud and AI services, which can lower barriers to entry and broaden access to advanced capabilities. Such openness, however, must be underpinned by trusted governance, with credible rules to safeguard privacy and intellectual property, ensure cybersecurity and strengthen operational resilience.
AI is unlikely either to end services offshoring or simply to accelerate it. Rather, it will make some tasks easier to coordinate across borders while eliminating the need to trade others. Countries will therefore need to take a more systemic approach, competing not only for individual tasks, but also for capabilities, business functions and intangible assets.
The central question in the next phase of globalization is not simply where a service is produced. It is how a problem is decomposed, which tasks are traded, which are automated, how the remaining work is governed, and who ultimately owns the knowledge and relationships created in the process.
The writer is president of the University of International Business and Economics.
附原文链接：https://www.chinadaily.com.cn/a/202607/27/WS6a66b347a310986e2b46766d.html
