Executive judgment

The World Trade Organization’s October 2026 update records a structural break. AI-enabling goods—principally the semiconductor, server and related equipment stack—accounted for 47% of year-on-year growth in the value of world merchandise trade in the first half of 2026. Their trade value rose 67% year on year; their share of world merchandise trade reached 14.8%, from roughly 7–8% during 2016–23.1

This is not merely an unusually strong technology cycle. It is a change in the composition of trade growth, investment and strategic dependency. The WTO lifted its 2026 merchandise-trade-volume forecast from 1.9% to 3.9%, while noting that the top ten traders accounted for about 85% of AI-enabling-goods exports and 80% of imports in 2025.1 The inference is straightforward: an AI-led expansion can increase global trade while distributing its bargaining power, learning effects and fiscal returns very unevenly.

Zwarte Peper’s position: governments should welcome rapid AI investment and open digital trade. But they should not confuse access to imported compute with national capability, or concentration with efficiency. The appropriate response is pro-trade, pro-diffusion and pro-competition: reliable market access for inputs, interoperable digital rules, contestable infrastructure, public capability to deploy, and targeted capacity-building where the market leaves countries permanently dependent.

1. The data points to an infrastructure boom, not a generalised recovery

The WTO attributes the revision in part to strong AI-related capital investment, while the same update reduced its 2026 services-trade-volume forecast from 4.8% to 3.3% amid transport and travel disruption. This distinction matters. The goods upswing is being driven by a specific physical stack: advanced chips, servers, storage, networking, power equipment and the trade services around them. It is not evidence that every economy or sector has received an equivalent productivity dividend.2

For public institutions, that changes the policy frame. AI cannot be treated solely as software regulation. Procurement, electricity, cloud concentration, customs reliability, cybersecurity, skills, public data and competition policy now determine whether AI can be used at all—and on what commercial terms.

2. Concentration is a trade problem before it becomes a model problem

High concentration in AI-enabling trade creates exposure at several levels. A government may be able to buy an application but remain dependent on a small number of foreign cloud regions, chip supply chains, model providers and technical standards. A domestic firm may export an AI-enabled service while absorbing volatile prices and restrictive terms upstream. Neither outcome is fixed by indiscriminate localisation mandates: those can raise cost, shrink choice and slow deployment.

Counterargument: a concentrated trade share is not, by itself, proof of abuse or durable market power. It can reflect comparative advantage, capital intensity, security requirements and legitimate economies of scale. The policy implication is therefore due diligence, not presumption: measure switching time, technical interoperability, price-setting power, access to critical inputs and the feasibility of credible alternative supply before choosing either intervention or reliance on the status quo.

The legal-commercial task is more precise. Public buyers should require portability, audit access, transition support and separable technical layers. Trade negotiators should prioritise predictable treatment for data, digitally delivered services and professional expertise, while preserving proportionate cybersecurity and public-interest safeguards. Competition authorities should examine control of essential cloud, distribution and compute access using evidence of switching cost and foreclosure—not slogans about national champions.

There is now a live institutional route for this work. The WTO’s Electronic Commerce Agreement is proceeding through interim arrangements: 68 members have undertaken them, and the agreement will enter into force for accepting members after 45 instruments of acceptance are deposited. Its legal pathway is therefore material, but incomplete; public leaders should distinguish a participating coalition’s interim route from multilateral incorporation into the WTO architecture. The implementation question is not whether to copy another jurisdiction’s rules, but how to translate interoperable commitments into domestic capability, safeguards and genuinely contestable market access.4

3. Development cannot be an afterthought to interoperability

UN Trade and Development’s 2026 report reaches the complementary conclusion: AI-related trade is a key source of dynamism, but its gains are highly concentrated, while developing economies face technological and policy barriers.3 That is an argument for institutional capability, not protection for its own sake. Countries need dependable connectivity, power, skills, trusted data institutions, research access and procurement systems that allow local adaptation. They also need participation in the standards and trade rules that determine what counts as a compliant service or secure cross-border transaction.

A rules-based order remains essential precisely because fragmentation can lock in hierarchy. But “open” cannot mean a system in which a small group supplies infrastructure and everyone else is confined to demand. Development-sensitive trade policy should aim to lower the cost of participation and increase the number of credible producers, adapters and evaluators.

4. Start with a service that makes trade rules usable

Brazil’s Ministry of Development, Industry, Trade and Services and Secretariat of Foreign Trade have deployed Lina, an AI assistant on the Siscomex single window. The WTO records that the service provides guidance grounded exclusively in official government sources; after its October 2025 launch it logged more than 2,500 sessions by June 2026, while users had previously waited up to four days for procedural answers.5

This is a useful implementation pattern, not proof that every border chatbot will reduce clearance times or costs: the WTO case study is descriptive rather than an independent impact evaluation. But it identifies a credible first deployment for lower-capacity administrations. Anchor an assistant to authoritative procedures, show the source and currency of each answer, route uncertainty and case-specific questions to officials, and measure whether the service reduces repeat enquiries without increasing erroneous guidance. That improves access before an administration takes on harder uses such as risk scoring or inspection prioritisation.

Commercial inference. The scarce asset is not a general-purpose chat interface. It is a maintained, legally current and interoperable public information layer connecting traders, customs, licensing agencies and a human remedy path. Building that layer also improves the data discipline needed for later automation.

5. A practical agenda for public leaders

  1. Measure exposure. Map critical AI inputs, cloud concentration, service dependencies and exit paths—not only model usage.
  2. Buy for substitutability. Use open interfaces, exportable records, documented evaluation evidence and transition obligations in consequential procurements.
  3. Keep trade rules usable. Reduce friction for legitimate hardware, services and expertise while maintaining clear, risk-based rules for security and data protection.
  4. Invest in diffusion. Direct public capability toward multilingual services, local evaluation, skills and sectoral adoption rather than prestige infrastructure alone.
  5. Test concentration empirically. Review prices, switching time, interoperability and access to essential inputs before choosing either laissez-faire or localisation.

The WTO’s update is not a recommendation for one industrial model. It is a warning against analytical lag. AI has become a material driver of trade statistics and commercial strategy. Policy must now catch up to its physical, legal and distributional consequences.

Authorities and methodology

  1. World Trade Organization, Global Trade Outlook and Statistics, October 2026 executive summary.
  2. World Trade Organization, “AI boom and trade resilience to Middle East crisis lift outlook, but gains to be uneven”, 8 October 2026.
  3. UN Trade and Development, “Global growth slows to 2.6%, and gaps widen between developing regions”, 9 October 2026; Trade and Development Report 2026: The geoeconomics of development.
  4. World Trade Organization, Agreement on Electronic Commerce, legal status and interim arrangements, checked 25 October 2026.
  5. World Trade Organization, “Brazil: Ministry of Development, Industry, Trade and Services, and the Secretariat of Foreign Trade”, AI for Trade in Action case study, checked 8 November 2026.
  6. World Trade Organization, World Trade Report 2026, executive summary, simulations on AI, checked 10 November 2026.

Editorial note. This is general research, not legal or investment advice. Figures and institutional status were checked on 11 November 2026.