Asia’s different routes to AI advantage and what Australia must learn

Artificial intelligence is often described as a race. That metaphor encourages governments to ask who is ahead, who is behind and how quickly their firms are adopting the technology.

But it obscures a more important reality: countries are competing in different parts of the AI economy and through very different national strategies.

AI is not a single technology. It is a global production system built across five interdependent layers: energy, semiconductors, digital infrastructure, models and applications. Economic value is created throughout this stack, but it is not distributed evenly. Some countries supply the chips, some host the computing infrastructure, some build the models, and others apply foreign technologies to existing industries. A country can benefit substantially from AI while capturing only a limited share of its commercial value.

A new AI Stack Competitiveness Index I have developed compares 48 economies: all OECD members and ten other major economies, including China, India, Singapore, Malaysia and Indonesia. The results reveal that Asia does not have one AI development model. It has several.

Singapore ranks second overall, behind only the United States. Its strength does not come from domestic scale. It comes from orchestrating an unusually dense combination of semiconductor activity, digital infrastructure, research, talent, multinational investment and internationally connected applications.

Singapore demonstrates how a small economy can compensate for limited scale by positioning itself as an indispensable node within global technology networks. Its advantage rests less on owning every part of the stack than on connecting multiple layers efficiently.

South Korea, ranked third, represents a different pathway. Its position is grounded in semiconductor capability, advanced manufacturing, infrastructure and strong corporate ownership. Unlike economies that predominantly import AI technologies, Korea possesses firms able to shape critical parts of the global production system.

China, ranked seventh, follows the scale-driven route. It combines a large domestic market with substantial industrial capacity, infrastructure, model development and domestic platforms. Its strategy is partly motivated by vulnerability: external restrictions on access to advanced chips and technologies have reinforced the perceived importance of controlling more of the stack domestically.

China’s experience shows why AI policy is increasingly inseparable from industrial policy, trade policy and geopolitics. Control over chips, compute and models can affect not only productivity but also national bargaining power.

India presents yet another configuration. Its overall ranking is much lower, reflecting limitations in infrastructure, semiconductor capacity and domestically captured investment. But its position in the applications layer is considerably stronger.

India’s talent base, digital public infrastructure, software capabilities and globally connected services firms provide a foundation for application-led advantage. Its opportunity may not be to reproduce the American or Chinese stack. It may instead become a major location for adapting, deploying and managing AI across firms and markets.

Malaysia offers a more specialised pathway through its semiconductor and electronics ecosystem. Indonesia’s potential lies further downstream, in the application of AI across a very large and rapidly digitalising market. Japan retains considerable industrial and research capability but faces the challenge of converting established technological strengths into leadership in the newest layers of the AI economy.

These differences matter for Australia.

Australia ranks 23rd in the index. Its position is neither that of a technological leader nor a country without meaningful capability. It possesses valuable foundations: energy resources, critical minerals, strong universities, trusted institutions and internationally competitive expertise in sectors such as mining, agriculture, education, health and professional services.

Its weaknesses are equally apparent. Australia depends heavily on foreign firms for advanced semiconductors, hyperscale cloud infrastructure, frontier models and digital platforms. As AI adoption accelerates, imports of chips, servers, cloud capacity, software and model services are likely to rise before Australia develops a comparable new wave of AI-enabled exports.

Australia could therefore become an exceptionally capable user of AI without becoming a significant owner of the AI economy.

The Asian comparison suggests that Australia should not attempt to recreate the entire stack domestically. Technological self-sufficiency would be prohibitively expensive and inconsistent with the scale of the Australian market. Passive dependence, however, is equally unsatisfactory.

The more credible strategy is selective stack-building.

Australia should identify the points where its existing capabilities intersect with emerging AI demand. Renewable-powered data infrastructure is one possibility, but attracting foreign-owned data centres is not sufficient. Their contribution should also be assessed through domestic procurement, workforce development, research partnerships, tax outcomes and the opportunities they create for Australian firms.

Critical minerals provide another opening, but exporting unprocessed inputs will capture only a small share of the value created further along the stack. Australia should connect its resource position with processing, advanced materials, energy systems and trusted supply-chain partnerships.

The largest opportunity may ultimately lie in specialised applications. Australia does not need to build the world’s largest general-purpose model to become internationally competitive in AI-enabled mining, agricultural technology, education, health, finance or professional services. But these applications must produce Australian intellectual property, firms, exports and customer relationships, not simply productivity gains for domestic users of foreign platforms.

Partnerships across Asia will be essential. Korea and Japan offer semiconductor and industrial capabilities. Singapore provides infrastructure, capital and ecosystem connectivity. India offers talent, services and application development. Australia brings energy, minerals, research, trusted institutions and sector expertise.

The future of AI in Asia will not be determined by a single national race. It will be shaped by countries occupying different positions across an increasingly interdependent production system. Australia’s task is not to dominate every layer, but to determine where it can become indispensable, and ensure that more of the resulting value remains at home.

Vikas Kumar is Professor of International Business in the University of Sydney Business School.