Press | GFTN

Geo-Economics in the Age of AI

Written by GFTN | Nov 18, 2026, 3:25:37 AM

Ravi Menon

Chairman, Global Finance & Technology Network

 

Opening Address, Singapore FinTech Festival 2026

Festival Stage, Singapore Expo

18 November 2026

 

Good morning, and welcome to the eleventh edition of the Singapore FinTech Festival.

Each year, we choose a theme that captures what is happening to finance. This year, the theme is rewiring.

The global financial system is being rewired by five forces at once.

  • Geo-economics: Contestation is redrawing the corridors along which trade and investment have flowed for decades.
  • Technology: Artificial intelligence is moving from advising to acting.
  • Capital: Money is being reallocated at scale, raising questions about whom it serves.
  • Talent: Skills are expiring faster than institutions can renew them.
  • Policy: It must keep pace with all of this, without compromising trust.

Rewiring is delicate work. The lights must stay on while the circuits change.

  • Payments must still clear, savings must still be safe, credit must still flow.
  • Done well, rewiring creates opportunity. Done carelessly, it creates fragility.
  • GFTN asked more than fifty leaders from around the world to tell us what each of these five rewirings means for the year ahead.
  • Their views are collected in one volume entitled The First Word, which we are launching … right now![1]
  • To all the contributors who shared their insights and convictions so generously: thank you. The first word of the Festival is yours!

This morning, I want to focus on one of the five rewirings, geo-economics, and on how it is being transformed by another, technology.

THE NEW GEO-ECONOMICS

The world has entered a new age of geo-economics.

  • Geopolitics pursues power through armies and alliances. Geo-economics pursues it through markets, money and machines.
  • Its toolkit includes coercive instruments: tariffs, sanctions, export controls, investment screening, payment restrictions.
  • But it also includes co-operative instruments: trade agreements, harmonised standards, supply chains, digital corridors.
  • The trouble is that the first set is being used far more than the second.
  • I spoke about this at some length four months ago.[2] Here's the QR code to that speech.

This new geo-economics is playing out along five flashpoints.

  • Trade: It is not just about comparative advantage. It is also about strategic positioning.
  • Investment: Inbound investments are screened for security risks. Outbound investments are screened for technology leakage.
  • Supply chains: What used to be designed for efficiency is now being reconfigured for resilience.
  • Technology: Strategic competition is creating bifurcated technology stacks.
  • Finance: It is not just about allocating capital. It is also about exercising power.

But while multilateralism has weakened; globalisation has not.

  • Trade restrictions have increased, but so have trade volumes.
  • Supply chains have not retreated, they have reconfigured.
  • The power of comparative advantage remains relevant.
  • The task is to manage interdependence, not to escape it.

We must shape a new globalisation: more resilient, more organic than the old one.

  • We need new partnerships:
    • plurilateral coalitions of the willing where multilateral consensus is lacking; and
    • public-private partnerships where government action alone is insufficient.
  • And where better than the Singapore FinTech Festival to forge these new partnerships?

A key factor to bear in mind as we seek new partnerships in the new geo-economics is artificial intelligence.

  • AI is arriving at precisely the moment the global system is being rewired.
  • It runs through all five flashpoints.
  • It raises the stakes in each geo-economic flashpoint, exposes the limits of the instruments used to fight them, and offers new tools for a constructive new globalisation.

AI RAISES THE STAKES IN THE NEW GEO-ECONOMICS

First flashpoint, trade. AI has turned export controls into the sharpest trade instrument of our time.

  • AI raises the stakes because the most contested goods in world trade today are its inputs: advanced chips, the tools that make them, and the compute they power.
    • Since October 2022, the US has been tightening controls on advanced chips and chipmaking tools bound for China.[3]
    • China has answered with instruments of its own: customs holds on licensed chips, guidance steering buyers to domestic suppliers, and export controls on rare earths suspended under last year's truce.[4]

But controls work best when control is near-total and substitutes are absent. Both conditions erode with time, and AI shortens the time.

  • Within three years of the first controls, Chinese labs had released open models close to the frontier, in one case trained at a fraction of the reported cost of their American counterparts.[5]
  • Domestic accelerators, a generation or two behind, supply a growing share of China's AI chip market, and improve every year the controls stay in place.

Second flashpoint, investment. AI has made compute the object of investment policy.

  • AI raises the stakes because the capital at play is now enormous and concentrated.
    • Hundreds of billions of dollars a year are flowing into data centres, much of it from a handful of technology companies.[6]
  • The instruments have escalated from screening to ownership.
    • In 2025, the US restricted American capital in Chinese AI, advanced semiconductors and quantum technologies.[7]
    • Months later, the US government took a 9.9% stake in Intel.[8]

An investment tool acts on who owns an asset. But it cannot act on what the asset needs.

  • The Netherlands took control of Nexperia, a Chinese-owned chipmaker, under a law never before used. Beijing blocked exports from Nexperia's Chinese plants, and within weeks carmakers in Europe and Japan warned of shortages.[9]
    • Nexperia's control changed hands; its dependence on its own Chinese packaging plants did not.

Third flashpoint, supply chains. AI rests on the most concentrated supply chain in the world.

  • AI raises the stakes because its own supply chain has almost no redundancy.
    • A handful of chip designers. One company making the most advanced lithography machines. One leading-edge foundry, on one island. Three hyperscale clouds.[10]
  • And the supply chain is physically entangled in ways a risk committee rarely sees.
    • A transaction powered by AI in Singapore may depend on a chip fabricated in Taiwan, a model trained in Texas, power generated from Qatari gas, and a cable running through the South China Sea.
    • Every link is a geo-economic exposure.

But stockpiles and subsidies buy time, not independence.

  • Chips are being warehoused like oil on both sides of the controls. But a stockpile of chips depreciates faster than a stockpile of oil.
  • Subsidised fabs replicate capacity. But they cannot replicate the ecosystem of suppliers, engineers and tacit know-how that grew up around the original.

Fourth flashpoint, technology. The bifurcation is leading to two separate stacks.

  • The semiconductor supply chain, once the most globalised in the world, is bifurcating.
  • AI raises the stakes because the divide now runs through every layer of the stack.
    • Chips: We have a US-led stack designed in California, fabricated in Taiwan, with Dutch lithography and Korean memory; and a China-led stack built on domestic accelerators and foundries.
    • Models: The US frontier models are mostly closed; you rent access. The leading Chinese models are open; anyone can download and run them.
    • Data regimes: America's is market-first, China's is security-first.

But a control on hardware cannot contain an idea that travels as a software download.

  • Export controls can stop a chip at the border. They cannot stop the AI model itself: in essence, a very large file that anyone can download, anywhere, in seconds.
    • It is a bit like restricting the export of printed books while the books travel freely online.
  • The more one side closes, the more attractive the other side's open models become to third countries that want capability without dependence.
  • And how distinct really are the two stacks?
    • Take a data centre in Southeast Asia. It runs on American chips, serving Chinese open models, under European-style data rules, all in one building.
    • Which stack does it belong to?

Fifth flashpoint, finance. AI raises the stakes of financial coercion, and accelerates the growth of the alternatives that blunt it.

  • Finance is the flashpoint that can be most damaging if it tips into fracture.
    • The freezing of Russia's reserves in 2022 was a watershed: the first time the coercive power embedded in the global financial system was deployed against a major economy.
  • The source of this power is the US dollar and its clearing system.
    • The US dollar accounts for more than half of official foreign reserves and of cross-border payments; it sits on one side of 9 out of 10 foreign exchange trades.[11]
    • And the dollar's dominance is extending through computer code. Nearly all stablecoins outstanding are dollar-denominated.

But while dollar dominance persists, its leverage is eroding at the margin.

  • Recent research published by the IMF shows that the power of economic coercion is nonlinear.[12]
    • The US and its allies control an estimated 80 to 90% of financial services in most countries.
    • At 95% dominance, a target of financial coercion has nowhere to go. But at 85%, it has options.
    • So an alternative does not need to win. It needs only to exist, to blunt the dominant power’s edge.
    • Russia's domestic payment rails helped mute the sanctions of 2022.
  • The researchers' prescription is neither de-dollarisation nor decoupling. It is targeted diversification in true chokepoints. And we are seeing more signs of this.
    • Many developing economies have put in place domestic payment rails: India's UPI, Brazil's Pix, Africa's PAPSS, ASEAN's payment linkages.
  • And as AI agents begin to route payments, the cost of switching to an alternative rail will fall sharply. I will come back to this.

Looking across the five flashpoints, a pattern emerges: every target of coercion adapts, and AI often shortens the time to adapt.

  • This does not mean that the instruments of geo-economics do not work. They do.
  • But their effects erode over time as alternatives reach a meaningful share.
  • AI accelerates that erosion, because it lowers the cost and shortens the time of building a substitute.

And it is why the constructive face of geo-economics matters more in the age of AI.

A CONSTRUCTIVE GEO-ECONOMICS, ENABLED BY AI

Plurilateral coalitions and public-private partnerships are the key vehicles for this constructive geo-economics.

  • Let me elaborate on one domain where this is playing out: interoperable payments.
  • The play is unfolding in three acts.

Act One: the cross-border payment connectivity journey is well underway.

  • The building block was laid in 2021 with the PayNow-PromptPay payment linkage between Singapore and Thailand, enabling cross-border transfers using only a mobile number.[13]
  • This created the blueprint for a series of bilateral ASEAN retail payment linkages.
    • Today, a tourist from Indonesia shopping in Malaysia can open his domestic banking app, scan the Malaysian merchant's QR code, and pay. The tourist is debited in Indonesian Rupiah, the merchant receives Malaysian Ringgit, and the settlement happens in seconds.
    • It bypasses correspondent banking networks, international card schemes, and the US dollar.
  • The goal now is multilateral scaling.
    • The central banks of five ASEAN countries and India, are building a hub-and-spoke model where each country’s national payment system connects to a central gateway and gains access to all others.
    • Called Project Nexus, it is targeting a live launch next year.[14]
  • Similar payment connectivity initiatives, like Project Lumina, are ongoing in Latin America[15].

Act Two: tokenisation is rewiring the settlement layer which lies beneath payment rails.

  • Digital money and tokenised assets are moving from pilots to production.
  • There is an opportunity here to develop a plurilateral standard, so that regulated digital money in one jurisdiction settles in another.
    • A common framework for tokenised bank liabilities could cut cross-border transaction costs by 12.5%, saving businesses more than US$50 billion a year by 2030.[16]
  • Project Guardian is Singapore's down payment. Its output is standards:
    • for fixed income with ICMA,
    • for foreign exchange with ISDA, and
    • a Guardian Wholesale Network to bring them to market.[17]

Act Three: agentic commerce may further integrate payments systems.

  • AI agents are beginning to search, compare, and pay on our behalf.
    • McKinsey estimates that agents could orchestrate US$3 to 5 trillion of global consumer commerce by 2030.[18] Much of this will be cross-border.
  • AI agents can help knit fragmented payment systems together, in two ways.
    • They hide the complexity. An agent can handle the currency conversion, message formats and compliance checks across different systems, so that the user experiences one seamless payment even when it crosses several rails.
    • They reward openness. Agents will favour rails that are open, cheap and transparent, and route around those that are not.

This is the next agenda for constructive geo-economics.

  • What Nexus is doing for payment systems, we now need to do for agentic AI: interoperable standards for agent identity and authorisation, so that an agent trusted in one jurisdiction can transact in another.

THE FOUNDATION: A SOCIAL LICENSE FOR AI

There is one more point I’d like to make before I close – and it’s the most important. It deserves a speech in its own right – maybe next year – but for now, let me simply state the case.

For AI to advance a constructive geo-economics, we must put people at the centre.

  • The populist backlash against globalisation was a verdict on domestic policy failure: the failure to share the gains of globalisation and support the displaced.
  • AI will face the same test. Let us not make the same mistake.

AI needs a social license before it can be widely adopted. That license rests on two things: jobs and wages.

  • AI does two things:
    • automation – which tends to displace jobs; and
    • augmentation – which tends to increase productivity and thereby wages and, in some cases, new jobs
  • The early evidence cuts both ways, and the junior worker sits at the centre of both.
  • Across the labour market, entry-level jobs show pressure first.
    • In the US, employment of 22-to-25-year-olds in the most AI-exposed occupations has fallen by about 13% relative to their peers since generative AI took off.[19]
  • But there is also evidence that within firms, AI lifts the least experienced workers most.
    • According to the first large field study of its kind, customer-service agents became about 15% more productive, and new hires gained the most.[20]
  • So the same tool can lift the junior worker or remove her. Which one happens is a choice. The employer decides, but market demand and public policy shapes the employer's choice.
  • Singapore is determined not to leave the spread of AI’s benefits to chance or to the market alone.

Get jobs and wages right, and AI earns its license to help rewire the global economy.

CONCLUSION

Let me conclude.

The new geo-economics is not the end of global connection. But the terms of connection have changed, and AI has changed them again.

A constructive geo-economics will not come from a grand bargain.

  • It will be built standard by standard, corridor by corridor, and now agent by agent …
  • by plurilateral coalitions and public-private partnerships that stay open …
  • and on a social license that gives it legitimacy at home.

None of this is pre-determined. But it is open to agency.

  • The instruments available to a small country are not tariffs and sanctions. They are standards, corridors, and trust.

That is Singapore's opportunity … and its responsibility: to be

  • a node where these forces converge,
  • a catalyst for the positive changes they demand, and
  • a convenor of the people who must make those changes together.

Many of those people are in this room. Let us get to work.

[1]Global Finance & Technology Network (2026). The First Word: The Five Rewirings of Finance, inaugural edition (GFTN), November 2026.

[2]Ravi Menon (2026). "The New Geo-Economics in a Quantum World", ST Lee Distinguished Lecture, S. Rajaratnam School of International Studies, Singapore, 14 July 2026. (https://gftn.co/the-first-word)

[3]US Department of Commerce, Bureau of Industry and Security. Interim final rules of 7 October 2022 (87 FR 62186), 17 October 2023 and 2 December 2024 on advanced computing and semiconductor manufacturing items; and final rule of 15 January 2026 revising licensing for certain advanced chips to China.

[4]Ministry of Commerce of the People's Republic of China, announcements of 9 October 2025 expanding export controls on rare earth items, suspended on 7 November 2025 for one year following the US-China understanding of 30 October 2025, and on 9 November 2026 extended to 10 January 2027. Reports of customs holds on Nvidia H200 chips and guidance to domestic buyers: Reuters, January 2026.

[5]DeepSeek-AI (2025). "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning", arXiv:2501.12948; DeepSeek-AI (2024). "DeepSeek-V3 Technical Report", arXiv:2412.19437, which reports a training cost of about US$5.6 million in GPU hours for the final run.

[6]International Energy Agency (2025). Energy and AI. Paris: IEA, April 2025; International Energy Agency (2026). Key Questions on Energy and AI. Paris: IEA, April 2026.

[7]US Department of the Treasury (2024). Provisions Pertaining to U.S. Investments in Certain National Security Technologies and Products in Countries of Concern (Outbound Investment Security Program), final rule implementing Executive Order 14105, 31 CFR Part 850, 89 Fed. Reg. 90398, issued 28 October 2024, effective 2 January 2025.

[8]US Department of Commerce and Intel Corporation (2025). Announcement of the US government's 9.9% equity stake in Intel, 22 August 2025.

[9]Government of the Netherlands, Ministry of Economic Affairs (2025). Order under the Goods Availability Act (Wet beschikbaarheid goederen) concerning Nexperia, 30 September 2025, announced 12 October 2025; Ministry of Commerce of the People’s Republic of China, export control notice concerning Nexperia, 4 October 2025; statements by the European Automobile Manufacturers’ Association (16 October 2025) and the Japan Automobile Manufacturers Association (23 October 2025).

[10]ASML is the sole supplier of extreme ultraviolet lithography systems; TSMC accounts for the large majority of leading-edge foundry output; Amazon Web Services, Microsoft Azure and Google Cloud account for roughly two-thirds of the global cloud infrastructure market (Synergy Research Group, 2025).

[11]International Monetary Fund (2026). "Currency Composition of Official Foreign Exchange Reserves (COFER)", IMF Data Brief, 27 March 2026; SWIFT (2026). Global Currency Tracker, January 2026 release; Bank for International Settlements (2025). "OTC Foreign Exchange Turnover in April 2025", Triennial Central Bank Survey 2025.

[12]Christopher Clayton, Matteo Maggiori and Jesse Schreger (2026). "Understanding Geoeconomics in a Volatile World", Finance & Development, International Monetary Fund, June 2026. The figures on control of financial services, the nonlinearity of coercive power, the 1% to 10% finding, the Russian example and the prescription of targeted diversification are all drawn from this article.

[13]Monetary Authority of Singapore and Bank of Thailand (2021). "Singapore and Thailand Launch World's First Linkage of Real-time Payment Systems", joint media release, 29 April 2021.

[14]Bank for International Settlements. "Project Nexus: enabling instant cross-border payments", bis.org; Nexus Global Payments (2026). "Nexus Global Payments Appoints Technical Operator", 9 February 2026 (PayNet-NETS joint venture; go-live targeted for 2027). Bank Indonesia joined as a full participant in February 2026.

[15] LAR launches Lúmina to advance the interconnection of instant payment systems in Latin America.” Press release, 24 September 2026.

[16]ISDA and Ant International (2025). Industry report on tokenised bank liabilities for FX settlement and cross-border payments under Project Guardian, 3 July 2025, citing Deloitte (2024) estimates of a 12.5% reduction in cross-border transaction costs, or more than US$50 billion in savings by 2030.

[17]Monetary Authority of Singapore. "Project Guardian", mas.gov.sg; MAS (2024). Guardian Fixed Income Framework (with ICMA) and Guardian Funds Framework, November 2024; Guardian Wholesale Network, announced November 2024.

[18]McKinsey & Company (2025). "The Agentic Commerce Opportunity: How AI Agents Are Ushering in a New Era for Consumers and Merchants", McKinsey QuantumBlack, October 2025.

[19]Erik Brynjolfsson, Bharat Chandar and Ruyu Chen (2025). "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", Stanford Digital Economy Lab Working Paper, August 2025: a 13% relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations, controlling for firm-level shocks.

[20]Erik Brynjolfsson, Danielle Li and Lindsey Raymond (2025). "Generative AI at Work", Quarterly Journal of Economics, Vol. 140, No. 2, pp. 889-942.