Deglobalisation, Globalisation 4.0 and the Future of Globalisation

Since roughly 2016 the standard description of the world economy has been that globalisation is going into reverse. The evidence is more awkward than that. Goods trade has flattened and capital flows never recovered their pre-crisis scale — but services trade is growing faster than output, data flows have expanded by orders of magnitude, and migration is at record levels. Integration is being reordered along strategic and regional lines rather than abandoned, and the difference matters.

What “Deglobalisation” Would Have to Mean to Be True

• Deglobalisation in its strict sense means a sustained fall in the cross-border movement of goods, capital, people and information, and a corresponding rise in the share of economic activity conducted inside national borders. It is a claim about direction and magnitude, not about mood.
• The 1914–1945 collapse is the only unambiguous case in the modern record, and it is the benchmark against which the present should be measured.
  • Trade as a share of output fell for three decades; capital markets closed; migration was choked off by quota laws; the gold standard broke.
  • Nothing in the current data is close to that, which is why the word is doing rhetorical rather than descriptive work.
• Slowbalisation is a much weaker claim and a much better-supported one. It describes a plateau rather than a decline.
  • Goods trade stopped rising as a share of world output after about 2008 and has moved sideways since, so integration is no longer deepening at the pace of the previous thirty years.
• Geoeconomic fragmentation is the IMF’s term and is different again: a policy-driven reallocation of integration along geopolitical lines.
  • States reorganise trade, investment, technology and payments around alignment rather than cost.
  • It says nothing about the total volume of trade. It is a claim about composition, which is why it can be true at the same time as trade grows.

Term	What it claims	Time reference	State of the evidence
Deglobalisation	Absolute, sustained decline in cross-border flows of all kinds	2016 onward, or 2008 onward	Not supported — services, data, migration all rising
Slowbalisation	Goods-trade intensity has plateaued; integration no longer deepening	From around 2008	Well supported on the goods measure alone
Geoeconomic fragmentation	Flows are being reallocated along geopolitical lines by policy	From 2018, sharply from 2022	Supported at the margin, strongest in high-technology sectors
Selective fragmentation	Some sectors and some relationships are decoupling; the rest are not	From 2018	Best fit with the disaggregated data
Re-globalisation	Integration is shifting its frontier from goods to services, data and climate	From around 2015	Supported by services and digital data; contested as a description of policy

The Metrics an Honest Test Requires

• Merchandise trade as a share of world output is the headline number and the one that produces the deglobalisation story. It is also the narrowest, because it measures the part of integration that matured first.
• Trade in commercial services — travel, transport, finance, business and professional services, computer services — is the fastest-growing component and is systematically under-measured, because services cross borders without passing a customs post.
• Cross-border capital flows and foreign direct investment, which behave very differently from each other: gross banking flows collapsed after 2008 and stayed down, while FDI has cycled without a comparable structural break.
• Migration stocks and remittance flows, the human dimension of integration, which almost never appear in commentary about deglobalisation and which point firmly the other way.
• Cross-border data flows, which are not in the trade statistics at all, which means the standard metric is structurally blind to the fastest-growing form of integration.
• Tariff and non-tariff measures — the policy input rather than the outcome — including subsidies, export controls, investment screening and local-content rules.
• The composition of trading partners rather than the volume of trade. Two economies can trade the same total while switching almost entirely who they trade it with, and that switch is what “fragmentation” actually names.

The metrics disagree, and that is the finding — not a problem with the data.

• Because the measures point in different directions, the question is genuinely contested rather than merely fashionable. A commentator who cites only merchandise trade will conclude that globalisation is ending; one who cites only services and data will conclude that nothing has changed. Both are reading a subset.

The Case for the Deglobalisation Thesis

The Two Structural Breaks: Goods and Capital

• Merchandise trade as a share of world GDP peaked around 2008 and has since moved sideways. The three decades before that saw trade grow at roughly twice the rate of output; since the crisis the two have grown at broadly similar rates, which means integration has stopped deepening on this measure.
• Cross-border capital flows fell very sharply after 2008 and never recovered their pre-crisis share of output — the strongest single datum for the thesis.
  • Gross financial flows had reached extraordinary levels by 2007, driven largely by European bank lending; the retrenchment that followed was permanent rather than cyclical.
  • The mechanism was regulatory as much as market-driven — higher capital requirements, resolution regimes and a supervisory preference for domestically matched funding made cross-border banking structurally less attractive.
• Global value chains stopped lengthening. The share of trade accounted for by intermediate goods crossing multiple borders — the technical signature of production fragmentation — flattened after the crisis, having risen steeply through the 1990s.
• A structural explanation supports the thesis independently of policy: China’s own import intensity fell as it moved up the value chain and began producing domestically the components it once imported. Less of the world’s largest manufacturing economy’s output now requires a border crossing.

The Policy Turn: Tariffs, Controls and Screening

• The tariff turn of 2018 was the first major breach in the post-war direction of travel. The United States imposed tariffs on a large share of Chinese exports, China retaliated, and neither side has substantially reversed. Predictability, not average tariff levels, was the casualty.
• The tariff escalation from 2025 went considerably further, resting on the International Emergency Economic Powers Act (IEEPA).
  • A global “reciprocal” band of roughly 10–41%, trafficking tariffs on China, Canada and Mexico, and country-specific additions including 40% on Brazil and 25% on India from 27 August 2025.
• That episode ended with a domestic legal check rather than a diplomatic one.
  • In Learning Resources, Inc. v. Trump on 20 February 2026 the US Supreme Court held that IEEPA does not authorise tariffs at all, the power to regulate commerce being distinct from the power to levy taxes. All IEEPA tariffs terminated on 24 February 2026.
  • Tariffs under Section 232 (steel, aluminium, vehicles, copper, timber, advanced semiconductors) and Section 301 were untouched and remain in force — the protection survived the instrument.
  • The substitution was immediate. A 10% global tariff under Section 122 of the Trade Act of 1974 took effect on 24 February 2026 and rose to 15% in March.
    • That authority is capped at 150 days and lapsed on 24 July 2026, when a Section 301 action replaced it.
    • Congress legislated no statutory replacement — the tariff wall was rebuilt from other delegated powers rather than from a fresh mandate.
• The analytically interesting fact is the sequence, not the rates: unilateral executive tariff-making was checked by a domestic court, not by the trading system.
• The stock of trade-distorting measures has accumulated rather than merely risen, because restrictions are added faster than old ones lapse.
  • Independent monitoring of national trade policies records interventions rising every year since 2008 and reaching a further peak in 2025.
  • The change in rationale matters as much as the count: more than half of recent United States industrial-policy measures explicitly cite national security or geopolitical competition rather than climate or competitiveness, and security-justified measures are far harder to reverse.
• Export controls have moved from a marginal instrument to a central one, and they run in both directions.
  • Controls on advanced semiconductors, semiconductor manufacturing equipment and design software were extended progressively from 2022.
  • China’s rare-earth and critical-mineral controls from April 2025, widened in October 2025 into a foreign direct product rule reaching any good containing Chinese-origin rare-earth material, showed the same weapon in the other hand.
• Investment screening regimes have spread across the OECD, converting inward foreign investment from a presumptively welcome flow into a security question. Screening now routinely covers semiconductors, critical minerals, biotechnology, ports, data centres and undersea cables.
• Subsidy competition has returned openly, which the multilateral system was specifically designed to discipline and now cannot.

The Rules That Stopped Being Enforced

• The WTO’s Appellate Body has been paralysed since December 2019 because the United States has blocked appointments to it, which means a losing party can appeal into a void and escape any binding ruling.
• The negotiating function failed earlier. The Doha Development Agenda, launched in 2001, is effectively dead; no comprehensive round has concluded since the Uruguay Round in 1994.
• MC13 in Abu Dhabi (February–March 2024) and MC14 in Yaoundé (26–30 March 2026) both closed without a ministerial declaration, MC14 being only the second ministerial held in Africa, after Nairobi in 2015 and ending with its chair conceding that members had run out of time.
• The surviving products are real but modest: the Trade Facilitation Agreement (in force February 2017) and the Fisheries Subsidies Agreement (2022). An institution whose output over three decades is customs procedure and fish is not setting the rules of the modern economy.
• The current trade numbers give the thesis a short-run argument as well.
  • Merchandise trade volume grew 4.6% in 2025, but the forecast is 1.9% for 2026 — 1.4% if Middle East conflict raises energy prices — recovering to 2.6% in 2027.
  • Goods and services together grew 4.7% in 2025, with 2.7% forecast for 2026.
  • North America is close to flat while Asia leads both import and export growth — reallocation rather than contraction.

The strongest version of the deglobalisation case is not that flows have collapsed but that the rules protecting them have.

The Case Against: What Has Gone On Growing

Services, the Fact That Breaks the Story

• Services trade has grown faster than goods trade for most of the past two decades and continues to do so. Everything else in this section is secondary to that.
• Commercial services trade grew 5.3% in 2025 and is forecast at 4.8% in 2026, against 4.6% and 1.9% for merchandise. Services are outgrowing goods in a year when goods trade is forecast to slow sharply.
• Digitally deliverable services — software, business process work, professional and technical services, financial and information services — are the fastest-growing component within that.
  • They require no shipping, clear no customs, and are largely immune to tariffs, which is precisely why the policy turn has not touched them.
  • They are also where the developing-country entry point now lies, since a services exporter needs bandwidth and skills rather than ports and industrial estates.
• Measurement lags reality badly here. Services delivered inside a multinational’s own internal accounts, or bundled into the price of a good, are recorded as neither export nor import, so the true services share of cross-border activity is higher than the statistics show.

The Integration the Trade Statistics Do Not Count

• Cross-border data flows have grown by orders of magnitude over the past two decades and continue to compound at double-digit annual rates, driven by cloud computing, streaming, platform services and now model training and inference.
• None of this appears in merchandise trade data, and only a fraction appears in services data. A search query, a design file, a diagnostic image and a training dataset crossing a border are integration events that the standard metric cannot see.
• The consequence for the argument is direct: the measure that produces the deglobalisation conclusion is systematically blind to the fastest-growing form of integration, so a conclusion drawn from it alone is not merely incomplete but biased in a known direction.
• The value carried is not marginal. Estimates from the mid-2010s already placed the economic value of cross-border data flows in the same order of magnitude as trade in physical goods, and the gap has widened since.

Value Chains Reorganised, Not Dismantled

• Global value chains have been reconfigured rather than shortened. The decomposition of production into stages performed in different countries continues; what has changed is where the stages sit and how much redundancy firms carry.
• Firms have moved from single-source to dual- and multi-source procurement, added inventory buffers, and duplicated qualification of suppliers. Each of those is a cost increase, not a withdrawal from cross-border production.
• Regionalisation is the dominant pattern within the reorganisation — production networks thickening inside Asia, inside North America and inside Europe rather than dissolving.
  • Intra-Asian trade in intermediates has continued to grow strongly, with ASEAN and India absorbing stages displaced from China.
  • The European Union remains the deepest example of regional integration anywhere and is not unwinding.

People and Money Still Crossing

• International migration is at a record level, with the global stock of international migrants at its highest recorded figure and rising. No period of genuine deglobalisation has coincided with record migration.
• Remittances to low- and middle-income countries reached about $685 billion in 2024, larger than foreign direct investment and official development assistance combined — a structurally significant flow that is stable precisely because it is driven by household ties rather than investor sentiment.
• India is the world’s largest recipient, at about $137 billion on the United Nations’ 2026 estimates, which makes the human dimension of integration a first-order matter for Indian macroeconomic stability.
• Foreign direct investment rose again in 2025 after two consecutive years of decline, reaching $1.6 trillion, up 6%.
  • The distribution was uneven: developing economies received $901 billion, up only 2%, against an 11% rise for developed economies, and the top 20 host economies took over 80% of global flows.
• The number of preferential trade agreements in force has gone on rising even as multilateral negotiation collapsed — 387 in force as of August 2026, from 640 notifications. Governments have not stopped liberalising; they have stopped doing it at the WTO.

Rerouting Mistaken for Reshoring

• The most common analytical error in this area is to read a change in bilateral balances as a change in dependence. They are not the same thing.
• Between 2017 and 2024, China’s share of direct United States imports fell by roughly seven percentage points — but its share of the value added embodied in those imports fell by only about two. Most of the apparent decoupling was goods taking a longer route.
• The gainers in United States import share over that period were the connector economies — Taiwan (+4.1 points), Vietnam (+3.7) and Mexico (+2.3) — many of whose exports embody Chinese components, materials and, increasingly, Chinese-owned production capacity.
• The implication cuts both ways.
  • For the deglobalisation thesis, it is damaging: the dependence that tariffs were meant to sever largely survives, relabelled.
  • For policy, it is expensive: the same input now reaches the same consumer through a longer chain, adding cost and emissions without adding security.
• Where genuine separation has occurred, it is sector-specific rather than economy-wide. Empirical work consistently finds that high-technology trade is far more sensitive to geopolitical distance than low-technology trade — which is the signature of selective fragmentation, not general retreat.

Trade did not shrink; it changed address, and a changed address was mistaken for a departure.

The Pandemic as the Moment Efficiency Lost Its Argument

What COVID-19 Actually Demonstrated

• The pandemic did not cause the turn away from hyper-globalisation — the tariff war, the Appellate Body’s paralysis and the capital-flow retrenchment all predate it. What it did was convert an economic argument about efficiency into a security argument about resilience, and that reframing is what stuck.
• Concentration risk was the central lesson. Supply chains optimised for cost had converged on very small numbers of suppliers and locations for critical inputs.
  • Semiconductors: a shortage that began in consumer electronics idled automobile assembly lines worldwide for the better part of two years, demonstrating that a chip worth a few dollars can halt a product worth tens of thousands.
  • Active pharmaceutical ingredients: a concentration of production in China and India that left every health system exposed to two jurisdictions’ export decisions.
  • Personal protective equipment: a commodity product whose global supply was concentrated enough that price and availability collapsed simultaneously.
• Just-in-time inventory proved fragile under correlated shocks. Buffer stocks had been eliminated as working capital; the system had been optimised for a world in which disruptions are local and uncorrelated, and 2020 was neither.
• Export restrictions on medical goods and food were imposed by dozens of countries within weeks. That governments reached for export bans reflexively, including among close allies, was the most politically formative fact of the episode.
• Vaccine nationalism followed — advance purchase agreements that locked up supply, export controls on vaccines and their inputs, and the failure of COVAX to deliver equitably or on time.
  • The dispute over the TRIPS waiver, settled at MC12 in June 2022 by a narrow vaccine-only text far short of what was proposed, is the enduring Southern grievance from the episode.

What It Accelerated

• Remote work and digitally delivered services expanded enormously, and did not fully revert. Work that had been assumed to require physical presence was performed across borders at scale, permanently widening the set of services that are tradable.
• E-commerce, cloud adoption and digital payments advanced by several years in several months, deepening exactly the form of integration that the trade statistics do not capture.
• The result is the pandemic’s defining paradox: it disrupted one kind of globalisation while accelerating another. Goods integration was interrupted; services and data integration were pushed forward.

Why the Reframing Endured

• Once supply-chain concentration is described as a national security vulnerability rather than a commercial risk, the policy instruments change. Security justifications attract subsidies, procurement preferences, export controls and screening — instruments that trade rules were designed to restrain.
• Security-based measures are also far more durable than crisis measures, because no government wants to be the one that removed a safeguard before the next shock.
• The pandemic thereby supplied the legitimating vocabulary for a policy turn whose causes were geopolitical and domestic-political. That is why it functions as the hinge of the period even though it caused very little of what followed.

Reshoring, Nearshoring, Friend-shoring and De-risking

The four terms are routinely used as synonyms and are not. They describe different geographies, different motives and different degrees of ambition, and the policy debate is much clearer once they are separated.

Term	What it means	Origin and moment	Honest assessment
Reshoring	Returning production to the home economy	Post-2008 manufacturing politics	Rare outside subsidised sectors; costs are the binding constraint
Nearshoring	Moving production to a geographically closer country	Post-pandemic logistics shocks	Real in North America; Mexico is the clearest case
Friend-shoring	Restricting supply chains to politically aligned countries	Articulated by the United States Treasury in 2022	Politically appealing, hard to define — who counts as a friend shifts
De-risking	Reducing dependence in specific critical sectors without general separation	EU formulation, 2023; endorsed by the G7 at Hiroshima	The formulation that has actually been adopted
China+1	Corporate strategy of adding a second production location alongside China	Emerged from firms, not governments	The most widely practised of the five

• Friend-shoring was articulated from 2022 as the deliberate confinement of supply chains to a large number of trusted partners, and it marked the moment when supply-chain geography became an explicit instrument of alliance policy.
  • Its weakness is definitional. Alignment is a spectrum, it changes with elections, and the countries with the necessary industrial capacity are not always the aligned ones.
• De-risking rather than decoupling was the European formulation of 2023 and became the G7’s language at the Hiroshima summit the same year. It concedes that separation from the world’s largest manufacturing economy is neither feasible nor desirable, and confines the ambition to identified critical dependencies.
  • The associated American formulation — a small yard with a high fence — makes the same distinction: intense restriction over a narrow set of technologies, normal commerce outside it.
  • The practical difficulty is that the yard keeps growing. Each new dual-use technology extends the perimeter, and there is no political mechanism for shrinking it.
• China+1 is the corporate version and is the one actually being executed at scale, because it requires no policy and no ideology — only a second qualified supplier.

Which of It Is Real

• The clearest genuine cases are those where a state has paid for the outcome. Semiconductor fabrication capacity is being built in the United States, Japan, Germany and India that would not exist on commercial logic alone.
• Electric-vehicle battery capacity has been relocated at scale, driven by domestic-content conditions attached to consumer subsidies — the most effective single instrument of the period.
• Pharmaceutical ingredients have attracted incentive schemes in several countries, though the cost differential remains large and the shift is partial.
• Against that, a great deal is relabelling. Announced investment substantially exceeds realised capacity in almost every category, announcements are counted repeatedly as they pass through political milestones, and construction timelines for complex facilities run to five years or more.
• The distinction to hold onto: an investment announcement is a political event; an operating plant is an economic one, and the gap between the two is where most of the reshoring narrative lives.

What It Costs

• The IMF puts the long-run output cost of trade fragmentation at between 0.2% and 7% of global GDP — the upper end roughly the combined annual output of Germany and Japan.
  • Some individual countries could lose 8–12% of output in a scenario that adds technological decoupling between blocs.
• Asia bears the largest share of the loss because it is the region most dependent on open trade, and low-income economies lose most in relative terms because they forgo technology transfer as well as market access.
• Duplicating supply chains is inflationary. Building second and third sources for the same input raises unit costs, and those costs are passed on where demand is inelastic and absorbed in margins where it is not.
• Resilience and efficiency are a genuine trade-off, not a free lunch. Redundancy is insurance; insurance has a premium; the honest policy question is how much of it to buy and for which inputs, not whether it is costless.
• There is a distributional point inside the cost. Consumers in advanced economies pay through prices, and producers in developing economies pay through lost market access — but only the first group votes in the countries setting the policy.

The Named Beneficiaries and the Limits of Substitution

• India, Vietnam and Mexico are the most frequently named gainers from diversification, with Taiwan, South Korea, Poland and Malaysia also identified in most assessments.
• Each has a distinct advantage: Vietnam proximity and cost; Mexico proximity and preferential access to the United States market; India scale, a large domestic market and a services base; Poland integration into European supply chains.
• The counter-point is the important one. China’s status as workshop of the world was built over decades through investment in ports, industrial parks, workforce training, component ecosystems and logistics networks — none of which can be replicated quickly.
  • A single assembly stage can be relocated in eighteen months. The supplier ecosystem beneath it cannot, which is why relocated assembly so often continues to import Chinese components.
  • The result is the rerouting pattern already described: the map of final assembly changes while the map of underlying capability does not.

Industrial Policy Returns to the Countries That Once Forbade It

• The most consequential policy development of the period is not tariffs but the open, large-scale return of industrial policy in the advanced economies — the same instrument those economies spent three decades telling developing countries not to use.
• The CHIPS and Science Act (2022) committed direct subsidy, loan guarantees and an investment tax credit to semiconductor fabrication, packaging and research on United States soil, with conditions restricting recipients’ expansion in China.
• The Inflation Reduction Act (2022) is the larger measure and the more consequential, because its clean-energy incentives are uncapped and conditioned on domestic and allied content — subsidy and trade instrument fused into one design.
  • Its effect on partners was immediate: the European Union treated it as discriminatory and responded with its own state-aid loosening rather than a dispute.
• The EU Chips Act (2023) aims to raise Europe’s share of global semiconductor production, alongside the Critical Raw Materials Act with its benchmarks for domestic extraction, processing and recycling, and a ceiling on single-country dependence.
• Japan and South Korea have committed large subsidies to fabrication capacity and materials, Japan explicitly to reverse three decades of decline in a sector it once led.
• India’s production-linked incentive schemes cover electronics, pharmaceuticals, telecom equipment, solar modules, batteries and more; the India Semiconductor Mission carries them into fabrication.

Investment Follows Strategy

• The cleanest single indicator that investment logic has changed is compositional. Strategic sectors — AI infrastructure, semiconductors, critical minerals and energy-transition technology — accounted for 44% of global greenfield project value in 2025, against 16% in 2020.
• That is not a change in the volume of investment but in its direction: nearly half of new project value is now committed in sectors selected for strategic rather than purely commercial reasons.
• It is also concentrated. The top 20 host economies took over 80% of global FDI flows in 2025, and strategic-sector investment is more concentrated still, because the inputs it needs — power, water, skilled labour, fiscal capacity — are unevenly distributed.
• This is the mechanism by which fragmentation becomes self-reinforcing: capital allocated on strategic grounds builds capacity in aligned locations, which makes future strategic allocation easier and commercial allocation harder.

Subsidy Competition Is Itself a Form of Fragmentation

• When several wealthy states subsidise the same sector simultaneously, the result is not more capacity where it is most efficient but duplicated capacity where the fiscal space is largest.
• The competition is unwinnable for developing countries. A state that cannot match a subsidy loses the investment regardless of its underlying cost advantage, which converts industrial location into a function of fiscal capacity rather than comparative advantage.
• It also erodes the rules from inside. Subsidies are the hardest category of measure for the trading system to discipline even when its dispute machinery works, and it does not currently work.
• The legitimacy cost is real and is felt acutely in the South: the countries that wrote the disciplines against industrial policy are now its largest users, having first ensured that others could not employ it during their own development.

Clustering: Blocs, Regional Agreements and the Countries That Refuse to Choose

• The likeliest description of where the world economy is heading is not isolationism but clustering into regions organised around spheres of influence, defined by a mixture of economic and geopolitical ties.

Arrangement	In force	Distinctive feature	Limit
USMCA	2020, replacing NAFTA	Tough rules of origin, local-content and wage rules, and a non-market-economy clause constraining members’ deals with China	Subject to review; exposed to unilateral United States tariff action
European Union	Continuous deepening	The deepest regional integration anywhere — single market, customs union, common regulation	Slow decision-making; enlargement stalled
Belt and Road Initiative	From 2013	Infrastructure-led connectivity across Asia, Africa, Europe and Latin America	Debt sustainability disputes; not a rules-based bloc
RCEP	January 2022	Largest agreement by population and output; unified rules of origin across Asia	Shallow tariff cuts; no labour or environment chapters
CPTPP	2018; UK acceded 2024	Deepest rules on digital trade, state enterprises, labour	Small combined market; the United States is not in it
IPEF	Launched 2022	Four pillars including supply chains and clean economy	No market access, so limited incentive to join or comply

• USMCA is the sharpest instance of a trade agreement used as a bloc-building instrument. Its rules of origin for vehicles are demanding, it attaches wage requirements to content, and its non-market-economy provision effectively gives each party a veto over the others’ trade agreements with China.
• The Belt and Road Initiative has been recalibrated rather than wound down. After the lending peak, official language shifted to smaller, “small and beautiful” projects — but the record does not match the rhetoric.
  • 2025 engagement reached record levels — roughly $128 billion in construction contracts and $85 billion in investment, with average deal size rising rather than falling.
  • The composition shifted decisively toward mining, metals and energy, with Africa’s share rising sharply — the pattern of a state securing critical-mineral and energy inputs, not of a retreat.

“The ‘small yet beautiful projects’ in the BRI propagated through official channels during COVID should be seen as bygone.” — Christoph Nedopil Wang

• RCEP matters less for its tariff cuts, which are modest, than for its single set of rules of origin across fifteen economies, which makes it administratively easier to build an Asian value chain than a cross-regional one.
• The CPTPP carries the deepest rules of any large agreement, and its accession queue — including applications from the United Kingdom, which acceded, as well as China and Taiwan — has turned membership into a geopolitical question rather than a technical one.
• The Indo-Pacific Economic Framework illustrates the limits of bloc-building without market access. It offers coordination on supply chains, clean energy, tax and anti-corruption, but no tariff concessions, which is why partners have treated it as a signalling exercise.

Hedging Is the Majority Position

• The clean two-bloc picture fails the evidence test. Most countries are hedging rather than choosing, maintaining security ties with one pole and trade ties with the other.
• Empirical work on trade reallocation finds fragmentation happening at the margin and asymmetrically, not as a clean split.
  • Greater geopolitical distance became significantly associated with lower trade flows only after 2022, and only once China is treated separately.
  • Advanced economies’ imports from China have held up or grown; China’s imports from advanced economies have fallen sharply, which widens imbalances without severing linkages.
  • The effect is concentrated in high-technology goods and largely absent in low-technology ones.
• The connector economies are the structural expression of hedging: countries that trade heavily with both poles and profit from the friction between them. Their existence is what prevents the split from completing.
• BRICS functions in this reading less as a counter-bloc than as a hedging platform.
  • Eleven full members after the 2024–25 enlargements, a partner-country tier created at Kazan in October 2024, and India in the chair for 2026.
  • It has a development bank and a stated preference for local-currency settlement, but no common external tariff, no common currency and no dispute mechanism — the things that would make it a bloc.

Where the Split Is Sharpest

• Critical minerals and the energy transition are where fragmentation is most advanced, because the inputs are geologically concentrated and the processing is more concentrated still.
• China’s dominance of rare-earth processing and permanent magnets turned a commercial position into leverage in April 2025, and the October 2025 extension to a foreign direct product rule showed how far that leverage reaches into third countries.
• The effects were immediate and uneven — automotive and aerospace production threatened across the United States, Europe and Japan, with yttrium exports to the United States collapsing from 333 tonnes to 17 tonnes in the eight months after the restrictions.
• A one-year suspension agreed in October 2025 stabilised flows without changing the underlying position: the dependence remains, and only its activation was paused.
• The deep irony of the energy transition sits here: it requires more globalisation to work — minerals, equipment, technology diffusion and finance all crossing borders at scale — while simultaneously generating the industrial competition that fragments those flows.

Financial and Monetary Fragmentation

Using the financial network coercively produced a predictable response: the states most exposed to it began building alternatives. The instruments are real, they are growing, and they remain far smaller than the system they are meant to hedge against.

• Alternative payment infrastructure is the first response. China’s Cross-Border Interbank Payment System (CIPS) has grown quickly — annual value reaching about $24 trillion in 2024 before flattening in 2025, participants past 1,600 — but it clears renminbi payments rather than replacing global messaging.
  • The distinction matters: SWIFT is a messaging network across every major currency, while CIPS is a clearing system for one. They are not substitutes.
  • The scale gap is decisive. The renminbi is around 3% of payments over SWIFT, against roughly 48% for the dollar and 24% for the euro.
• Central-bank digital currencies and cross-border pilots are the second. Project mBridge, linking several central banks on a shared platform for wholesale settlement, is the most advanced; the Bank for International Settlements withdrew from it in 2024, leaving the participants to continue independently.
• Local-currency trade settlement is the third, and the most widely adopted because it requires no new technology. India’s rupee vostro account arrangements allow trade invoicing and settlement in rupees with a growing list of partners, and comparable arrangements exist across Asia, the Gulf and Latin America.
  • The constraint is the same everywhere: a surplus partner accumulates a currency it cannot freely spend, which caps how far bilateral settlement can go without full convertibility.
• Bilateral currency swap lines between central banks have proliferated, providing crisis liquidity outside the dollar system for participants.
• Gold accumulation by central banks is the fourth and most revealing, because it is a vote against counterparty risk rather than for any particular alternative.
  • Purchases ran above 1,000 tonnes for three consecutive years and reached 863 tonnes in 2025 — down 21% year on year but still far above the 2010–2021 average of 473 tonnes.
  • Poland (102 tonnes), Kazakhstan (57), Brazil (43) and Azerbaijan (38) led — a list notably not confined to sanctioned or aligned states.

The Sober Assessment

• The dollar’s share of allocated reserves has drifted down slowly over two decades, and no alternative has emerged with the depth, liquidity and legal infrastructure to replace it. At 57.1% it remains far ahead of the euro at 20.0%, with the renminbi in low single digits.
• Reserve currency status rests on things that cannot be legislated: a deep market in safe assets, free convertibility, an independent judiciary willing to enforce contracts against the issuing state, and open capital accounts.
  • China has chosen not to meet several of those conditions, because doing so would mean surrendering capital controls that serve other objectives. The renminbi’s limits are a policy choice, not an accident.
  • The euro area has the legal infrastructure but lacks a single deep safe asset of sufficient size.
• The honest formulation is that the movement away from the dollar is real, incremental and defensive — hedging against being cut off, rather than a bid to replace the system. Predictions of imminent displacement have a long record of being wrong, and the underlying data show erosion rather than replacement.
• The consequential fragmentation in finance is therefore not currency substitution but the segmentation of payment and settlement infrastructure, which raises transaction costs and reduces transparency for everyone without shifting the monetary hierarchy.

From Goods to Data: The Frontier of the Next Phase

Baldwin’s Third Unbundling

• Richard Baldwin offers the most useful account of what globalisation becomes next, by reading its history as a sequence of unbundlings — separations that falling costs made possible.
  • The first unbundling separated production from consumption, when cheap transport allowed goods to be made far from where they were used.
  • The second unbundling separated the stages of production from each other, when cheap communication allowed a factory’s functions to be split across countries — the origin of global value chains.
  • The third unbundling separates labour services from labour’s physical location: the worker stays put and the work crosses the border.
• Telemigration is his term for it — a professional in one country performing work inside another country’s economy in real time, without migrating and without a visa.
  • The pandemic supplied the proof of concept at scale, normalising remote delivery in exactly the occupations that had assumed physical presence.
  • Machine translation removes the language barrier that limited it, which is why the addressable pool is far larger than the English-speaking one.
• Artificial intelligence extends the same logic. Where the second unbundling exposed manufacturing workers in rich countries to competition from abroad, the third exposes professional and service workers — and it does so faster, because relocating a service requires no capital expenditure.
• The political implications run directly into the backlash already visible over goods trade. The constituency with political voice in advanced democracies is the one now facing the adjustment, which is why the coming contest over services and data will be sharper than the one over manufacturing.

Data Localisation and the Regulatory Contest

• Data localisation — requiring that data be stored, processed or mirrored within national borders — is the characteristic protectionism of digital globalisation, and it is justified on three quite different grounds.
  • Privacy: that citizens’ data should sit under a jurisdiction that protects it, which is the European framing.
  • Security and law-enforcement access: that the state should be able to reach data held about its territory, which is the framing most governments actually act on.
  • Industrial policy: that domestic processing builds a domestic data industry, which is rarely stated openly and often the operative motive.
• The European Union’s General Data Protection Regulation exports its standard through adequacy decisions and contractual clauses, making market access conditional on regulatory convergence — the clearest case of a large market setting global rules unilaterally. The Data Act extends the same logic to industrial and machine-generated data.
• China’s regime is the most restrictive, combining cybersecurity, data security and personal information laws with security reviews for outbound transfers and mandatory localisation for defined categories.
• India’s Digital Personal Data Protection Act 2023, with its rules notified in 2025, adopted a negative-list approach — transfers are permitted except to countries the government notifies as restricted — which is considerably more liberal than the strict localisation many expected.
  • Sectoral localisation remains: payment system data must be stored in India under the central bank’s 2018 mandate, and comparable rules apply in other regulated sectors.
  • The debate is unresolved rather than settled, and turns on whether India treats data primarily as a national resource to be processed domestically or as an input to a services export sector that depends on being able to move it.
• The cumulative effect of these regimes is regulatory fragmentation of the internet — not a splintered network, but a set of overlapping compliance zones that raise the fixed cost of operating globally and therefore favour the largest incumbents.

Taxing the Digital Economy

• The digital economy broke the tax rule that profits are taxed where a firm has physical presence, because a platform can serve a market intensively with no presence in it at all.
• The OECD/G20 two-pillar solution was the multilateral answer: Pillar One reallocating a share of the largest multinationals’ profits to market jurisdictions, Pillar Two setting a 15% global minimum effective tax rate.
• The outcome so far is a split verdict. Pillar Two is being implemented — over sixty jurisdictions with legislation drafted or adopted, and the European Union enforcing from 2024 — while Pillar One has stalled, its multilateral treaty unsigned well past its deadline.
• The United States has implemented neither pillar and has threatened retaliation against countries applying the minimum tax to American firms.
  • In January 2026 it secured a side-by-side arrangement from the OECD/G20 Inclusive Framework, letting United States-parented groups elect out of the income inclusion rule and the undertaxed profits rule for fiscal years from 1 January 2026.
  • The effect is the global minimum tax preserved in form and weakened in reach — the standard outcome when a rule binds everyone except the state large enough to refuse it.
• The consequence is the spread of unilateral digital services taxes and the trade retaliation they attract. Tax is now a trade instrument, which is precisely what the two-pillar process was designed to prevent.

30 March 2026: The Moratorium Lapses

• The WTO moratorium on customs duties on electronic transmissions, in place since 1998 and renewed at every ministerial or General Council since, lapsed on 30 March 2026 at the close of MC14 in Yaoundé — the first lapse in its history and the single most concrete instance of digital fragmentation to date.
• The dispute behind it is old and substantive rather than procedural.
  • Proponents argued that duty-free electronic transmission is what allowed digital trade to grow, and that the alternative is customs assessment of intangibles that cannot be inspected at a border.
  • India and South Africa led the opposition on three grounds: that the scope of “electronic transmissions” was never defined and expands as more goods dematerialise; that developing countries forgo tariff revenue as physical imports become downloads; and that the moratorium forecloses industrial policy in digital sectors.
  • The argument that carried most weight politically was the third one restated as fairness: the moratorium gave large foreign technology firms a tax advantage over domestic competitors in developing markets.
• What replaced it is a plurilateral. Sixty-six members representing about 70% of world trade are proceeding with an E-Commerce Agreement covering digital trade facilitation, consumer protection, electronic authentication and a commitment not to impose duties on electronic transmissions among themselves.
• This is the structural shift in miniature: a universal rule binding all 166 members replaced by a club rule binding sixty-six of them. Certainty is preserved for members and lost for the rest.
• The practical effect in the short run is limited — most countries lack the machinery to levy duties on electronic transmissions and few intend to — but the precedent is not limited. A rule that lapses once can lapse again, and the moratorium’s permanence was itself the thing being defended.

Compute as the New Chokepoint

• The decisive scarce input of the next decade is AI compute: advanced accelerators, the equipment to make them, and the power and cooling to run them.
• Control over it is being exercised through export licensing rather than tariffs, which is the pattern of the period — restriction by permission rather than by price.
• The policy has oscillated. Controls on advanced accelerators were progressively tightened from 2022, then partially relaxed and codified in January 2026 into a case-by-case licensing regime for the most capable chips, with certification requirements, end-user verification and a defined list of countries of concern.
• The reciprocal move is domestic substitution: restricted states accelerate indigenous design and fabrication, which over time converts a temporary denial into a permanent second ecosystem — the opposite of the intended effect.
• An AI governance divergence is emerging alongside it — the European Union regulating by risk category, the United States relying on sectoral rules and executive action, China licensing models and content directly, and most other states writing nothing. Divergent AI rules will do to services trade what divergent product standards did to goods trade.

After Multilateralism: What Is Actually on Offer

• The WTO is not dead but it is doing only one of its three jobs. It administers existing agreements and monitors policy; it has stopped negotiating and it can no longer reliably adjudicate.
  • 166 members after the accession of Comoros and Timor-Leste in February 2024, which means the institution is still worth joining.
  • The Appellate Body has been unstaffed since December 2019; the stopgap is the Multi-Party Interim Appeal Arbitration Arrangement (MPIA), with 61 participants covering roughly 60% of world trade by 2026.
  • The MPIA is a revealing institution: a plurilateral workaround that reproduces the multilateral function for those who opt in, which is the shape of everything else on this list.
• MC13 (Abu Dhabi, 2024) and MC14 (Yaoundé, 2026) both ended without a ministerial declaration. At MC14, 129 members backed bringing the Investment Facilitation for Development agreement inside the WTO and 165 of 166 supported its incorporation — and it still remains outside, because consensus means one objection is decisive.

Alternative	How it works	Advantage	Structural objection
Plurilateralism and joint statement initiatives	Subsets of members negotiate binding rules among themselves	Moves at the pace of the willing; can extend benefits on an MFN basis	Marginalises those not in the room; rule-taking without rule-making
Mega-regionals (RCEP, CPTPP, EU agreements)	Deep agreements among large groupings	Deals with services, data, standards	Fragments rules of origin; excludes non-members by design
Bilateral agreements	State-to-state, proliferating rapidly	Fast, tailored, politically saleable	Bargaining power decides terms; a noodle bowl of rules
Minilateral clubs (G7, G20, BRICS, Quad)	Coordination without treaty obligation	Flexible; can act on new issues quickly	No enforcement, no universality; membership is self-selected
Issue coalitions (climate clubs, mineral partnerships)	Narrow, functional, often sectoral	Matches the problem’s shape	Proliferation itself is fragmentation

• The trade-off to state plainly: plurilateral clubs move faster and can free-ride on most-favoured-nation treatment, but they marginalise members who are not in the room. That is the structural objection from smaller economies, and it is not answered by pointing at efficiency.
• There is a serious counter-argument that the order is more durable than the institution.
  • Ikenberry contends that rising powers seek greater authority within the liberal order rather than an alternative to it, because openness is what made their own rise possible.
  • On that reading the road to modernity runs through the existing rule-based order rather than away from it.
• The rejoinder is that an order can decay without ever being replaced. No competing organising logic has emerged, but an order whose rules are unenforced and whose disputes go unheard erodes by disuse rather than by defeat.

Globalisation 4.0 and the Competing Futures

• Globalisation 4.0 was the World Economic Forum’s 2019 Davos theme, associated with Klaus Schwab: integration with safeguards, adapted to an age of robotics, the internet of things and artificial intelligence, and to the social inequalities and political tensions those technologies produce.
• Its substantive claim is that the institutions of the trading order were built for goods and are now asked to govern services, data and technology — which a system designed in 1947 and updated in 1994 cannot do without redesign.
• The associated policy vocabulary — resilient, inclusive, “building back better” — signals a tempering of pure market liberalism rather than its abandonment: efficiency balanced against security and equity, with the balance left undefined.

Vision	Core claim	Named with	Evidence for it
Thinner globalisation	Fewer, shallower international rules; restored national policy space to rebuild domestic bargains	Rodrik	Strong — industrial policy’s return is exactly this
Re-globalisation	Integration continues but on new terms, centred on services, digital trade and climate	Reform economists and the WTO’s own agenda	Strong on flows; weak on rules
Two or three blocs	A United States-led and a China-led sphere, with a European third	Strategic commentary	Weak — hedging dominates, connectors profit
Regionalisation	Production and rules thicken within regions, thin between them	Widely held	Strongest single fit with the data
Status quo with friction	Same system, higher costs, more uncertainty	Market analysts	Fits the aggregate numbers well

• On present evidence the regionalisation and thinner-globalisation readings fit best, and they are compatible: deeper rules inside regions, thinner rules between them, with services and data continuing to integrate globally because no one has built the machinery to stop them.
• Rodrik’s diagnosis is the sharpest of the reform arguments.
  • The post-1990 model placed multinational firms in the driver’s seat and prioritised market integration over shared prosperity and economic security, achieving integration at the cost of domestic disintegration.
  • His conclusion is not reversal but a thinner globalisation leaving states room to rebuild their own social contracts — a complete collapse being unlikely, and a return to the status quo ante impossible.
• The counter-position, associated with Catherine Mann, is that the problem was never too much globalisation but too little of the domestic policy that makes it bearable — and that retreating from integration while leaving distributive policy unreformed is the worst of both.

The Climate Dimension

• The energy transition is where the contradiction is sharpest: it requires deeper globalisation — critical minerals, equipment, finance and technology diffusion at scale — while generating new protectionism in the industries it creates.
• The European Union’s Carbon Border Adjustment Mechanism entered its definitive phase in January 2026, requiring importers of cement, iron and steel, aluminium, fertilisers, electricity and hydrogen to surrender certificates matching the embedded emissions of what they bring in.
• Its logic is defensible — a carbon price is undermined if production simply relocates to unpriced jurisdictions — and its effect is nonetheless a trade barrier applied by the rich to the poor, since exporters must meet a standard set by an economy that decarbonised after industrialising.
• It is read across the Global South as green protectionism, and the objection has a legal dimension too: a unilateral measure with extraterritorial effect, introduced while the dispute-settlement system that might have tested it is paralysed.
• Green industrial policy compounds this. Domestic-content conditions attached to clean-energy subsidies make the cheapest path to decarbonisation illegal to subsidise, which raises the cost of the transition in order to distribute its industrial benefits domestically.

India Between the Blocs

India is the most interesting single case in this argument because it is simultaneously a beneficiary of fragmentation and a loser from the erosion of multilateral rules. Diversification away from China sends it investment; the collapse of enforceable trade law removes the protection that a middle-sized economy most needs against larger ones.

How Much of China+1 India Has Actually Captured

• The headline case is electronics. iPhone exports from India reached about ₹2 trillion in FY26 — roughly $22 billion, three-quarters of total smartphone exports of about $29 billion — from a base near zero five years earlier.
• The production-linked incentive schemes did the work, and the electronics scheme was the clearest success among them. Two contract manufacturers account for most output, with an ecosystem of over forty component suppliers.
• The honest qualification is domestic value addition, which remains low. India’s electronics exports rest heavily on imported components, and on this measure India sits near the bottom among large manufacturing economies.
  • Assembly relocated; the supplier base largely did not, which is the rerouting pattern seen from India’s side of the transaction.
  • Policy has recognised this — a component-manufacturing scheme and stated targets to raise mobile-phone value addition toward 35–40% — but the deep ecosystem takes a decade, not a budget cycle.
• The India Semiconductor Mission has approved ten projects worth about ₹1.6 lakh crore, with the first commercial fabrication plant at Dholera expected to produce from 2028. India today operates mainly in assembly, testing and packaging and imports the overwhelming majority of its chips.

The Agreement Push, and the Bloc It Stays Out Of

• India has moved from FTA scepticism to an unusually rapid bilateral push.
  • UAE CEPA and Australia ECTA (2022) in force; EFTA TEPA in force October 2025, carrying a $100 billion investment commitment — the first investment target India has secured in a trade treaty.
  • India–UK CETA in force 15 July 2026; the India–EU agreement concluded 27 January 2026 and in ratification.
  • Oman (December 2025), New Zealand (April 2026), and Canada terms of reference (March 2026).
• India remains outside RCEP, having walked away in November 2019 over import surges, an inadequate ratchet on tariffs and the absence of meaningful services access.
• The pattern is coherent rather than contradictory: bilateral agreements with partners whose exports do not threaten Indian manufacturing, and abstention from the one bloc containing China.
• The cost is real. Staying out of RCEP means staying outside the single set of rules of origin that makes an Asian value chain administratively cheap to build, at exactly the moment India is trying to join those chains.

Strategic Autonomy in an Economic Register

• Strategic autonomy is the framing India applies to the choice between blocs, and in economic policy it means hedging: Quad supply-chain work and a technology partnership with the United States alongside BRICS membership, the SCO, continued Russian energy purchases and a chairship of BRICS in 2026.
• The hedge has been tested. The 25% additional United States tariff from August 2025 showed how exposed the position is when one pole treats another relationship as grounds for punishment; its termination in February 2026 came from an American court, not from a negotiation India won.
• CBAM is the clearest current cost. Iron and steel and aluminium are significant Indian exports to the European Union and sit squarely inside the covered sectors, so the definitive phase from January 2026 imposes a carbon cost on precisely the goods where India competes on price.
  • India’s position is that CBAM is a unilateral measure with extraterritorial effect that ignores common but differentiated responsibilities, and it has pressed the issue in the EU trade negotiation rather than at a dispute panel that cannot hear it.
• On digital trade, India has been the most consistent opponent of the e-commerce moratorium for two decades, on revenue, scope and policy-space grounds.
  • Its lapse on 30 March 2026 is therefore a diplomatic success — and an ambiguous economic one, since India is a very large exporter of digitally delivered services and would be badly hurt if duties on electronic transmissions became general practice.
  • India is also outside the plurilateral E-Commerce Agreement, which means the rules for digital trade among two-thirds of world trade are being written without it.
• Atmanirbhar Bharat is best read not as autarky but as import substitution in strategic sectors combined with export ambition — tariff increases and incentives in electronics, defence and pharmaceuticals alongside the fastest bilateral trade-agreement programme in India’s history. The two are only contradictory if the policy is misdescribed.

Conclusion

In roughly seventy-five years the world moved from tight capital controls and fixed exchange rates, through a freewheeling global market ideology, into a more guarded global economy with power politics back on the agenda. The trajectory of globalisation is not linear, and the current phase is a reconfiguration rather than a reversal — more regional, more digital, more securitised, and fitted with more guardrails than the period it replaced.

The distinction matters because the two descriptions generate opposite errors. Policy built on the assumption that globalisation is ending will over-invest in self-sufficiency, forfeit the gains still available in services and data, and be wrong about where growth comes from. Policy built on the assumption that nothing has changed will leave critical dependencies unhedged and be wrong about where risk comes from. The accurate description is selective fragmentation, and getting it right is the whole of the analytical work.