1. Will AI favour the American or Chinese model of capitalism?
2. Will AI reward frontier innovation or large-scale deployment?
3. Will China dominate AI the same way it came to dominate EVs?
4. If China continues building industrial capacity and exporting aggressively as a competitive imperative — to gain profits, valuations and cheaper capital for the next cycle of innovation — what does this imply for Asean countries?
See also: The washing machine that explains China
But behind all four questions lies a much larger issue. AI may ultimately become one of the most transformational technologies in human history. Not merely because it automates tasks, but because it increasingly targets intelligence itself — the very capability that traditionally differentiates human beings economically.
Unlike previous industrial revolutions, AI does not merely replace muscle. It increasingly augments, replicates and potentially replaces cognition. AI is simultaneously frontier research, capital market, compute infrastructure, semiconductors, energy systems, data, industrial deployment, geopolitical and, increasingly, a civilisational challenge. That means AI may not likely fully favour either the American or Chinese system exclusively. Different phases of AI may reward different structural strengths.
Yet before discussing that, we must acknowledge that neither American nor Chinese represents a pure model. America still possesses enormous industrial and infrastructure capabilities. China increasingly produces world-class research, software firms and frontier AI talent. The distinction is not absolute superiority, but what each system structurally rewards more efficiently and at larger scale.
See also: Why China built capability and Japan preserved wealth
Phase 1: Frontier AI strongly favours American system
The current frontier AI overwhelmingly resembles classic American capitalism. Frontier AI requires massive risk capital, expensive long-duration research, elite global talent, strong intellectual property monetisation, deep software ecosystem, compute concentration, and capital markets willing to finance uncertain outcomes at extraordinary valuations.
That aligns closely with the American model described in our earlier article. OpenAI, Nvidia, Google, Anthropic, Meta and Microsoft are not just technology companies. They are products of a financial system capable of monetising future expectations at enormous scale — the US capital markets.
Nvidia itself illustrates this dynamic. Its valuation becomes strategic infrastructure. High valuations lower capital costs. Cheap capital funds supply chains, compute clusters and energy contracts. That attracts developers and customers, which deepens ecosystem dominance that in turn reinforces valuation.
This is classic American capitalism: the monetisation of future technological dominance before profits fully materialise. But AI will not likely remain purely a frontier software problem forever.
Diminishing returns question may determine the winner
The single most important unresolved economic question in AI may ultimately be this: Does frontier AI itself eventually face diminishing marginal returns?
For more stories about where money flows, click here for Capital Section
Historically, almost every major technology eventually does.
The early breakthroughs are revolutionary: the first aircraft, the first semiconductors, the internet, smartphones, search engines and cloud computing. But over time, improvements often become smaller, more incremental, more expensive and harder to commercialise economically.
Aircraft speeds eventually plateaued. Smartphone innovation slowed. Moore’s Law weakened. Even internet search quality matured.
AI may eventually encounter similar trajectories. Already, frontier models require extraordinary compute, enormous energy, vast datasets and escalating capital expenditure to achieve increasingly incremental gains. If GPT-8 only becomes marginally better than GPT-7, and GPT-9 only marginally better again, then the economics of AI changes profoundly.
When models become “good enough” for most commercial uses, the strategic question shifts from “Who has the smartest model?” towards “Who can deploy AI across the economy fastest, cheapest and at largest scale?”
And when that transition happens, it structurally favours China.
Phase 2: AI may increasingly become an industrial deployment problem
If frontier AI eventually experiences diminishing marginal returns, then AI increasingly becomes an infrastructure problem, an industrial deployment problem, a robotic problem, an energy problem, a manufacturing problem and a physical economy integration problem.
This is where China becomes extraordinarily competitive. China’s strengths are not just low-cost manufacturing. More critically, China excels at industrial deployment, ecosystem coordination, supply-chain integration, hardware optimisation, manufacturing scale, infrastructure execution and fast mass adoption. These are precisely the characteristics that enabled China today to dominate solar, batteries, electric vehicles (EVs), consumer electronics, telecom equipment, drones and, increasingly, robotics.
The West still tends to frame AI primarily as frontier models, scientific breakthroughs, large language models and software intelligence. But we think AI may increasingly industrialise. Eventually, the enormous investments in AI require economic returns. That means AI must move beyond demos and chatbots into factories, logistics, robotics, infrastructure, healthcare, transportation and the broader physical economy.
And China is already positioning AI as foundational industrial infrastructure. Huawei, Alibaba, Tencent and numerous industrial firms are integrating AI directly into manufacturing, logistics, surveillance systems, smart cities, vehicles and appliances.
China is simultaneously expanding energy generation, robotics adoption, semiconductor capabilities and hardware manufacturing ecosystems.
America currently dominates:
• Frontier models;
• AI chips;
• Hyperscale cloud ecosystems;
• High-margin software monetisation; and
• Global developer platforms.
China may increasingly become exceptionally competitive in:
• AI deployment;
• AI-enabled manufacturing;
• Industrial robotics;
• Hardware integration;
• Cost-efficient AI solutions; and
• Physical economy integration.
EV analogy matters, but only partially
The EV story illustrates how the industrial ecosystem compounds.
Initially, Western firms dominated technology, perception and branding and China lagged. Then China subsidised deployment, intensified domestic competition, scaled manufacturing, compressed costs, deepened supplier ecosystems and accelerated industrial diffusion. Eventually, Chinese EVs became globally competitive, cheaper and increasingly technologically sophisticated.
AI could also evolve similarly in humanoid robotics, Edge AI, smart manufacturing, autonomous systems, AI appliances, industrial automation and AI-enabled hardware. But AI differs from EVs in one crucial way. AI exhibits far stronger winner-takes-most dynamics. AI models improve with scale, data, compute, users and developer ecosystems. Platform dominance compounds globally.
Unlike EVs, AI also depends heavily on frontier semiconductors, advanced lithography, hyperscale compute infrastructure and software ecosystems.
Taiwan Semiconductor Manufacturing Co (TSMC), ASML, Nvidia’s Compute Unified Device Architecture (CUDA) ecosystem and hyperscale cloud infrastructure together form a deeply entrenched technology stack that is far harder to replicate than EV manufacturing. This means frontier AI may remain more defensible for much longer.
The key uncertainty for investors is therefore: Will AI eventually stabilise into an industrial deployment problem — like solar panels, batteries and EVs? Or will AI remain a continuously resetting frontier race where breakthrough innovation compounds faster than manufacturing scale? This single question determines whether America or China gains the longer-term advantage.
Our view is that timing matters. America likely dominates near-term frontiers. China may gain increasing advantages if AI industrialises broadly across the physical economy. But neither outcome is inevitable. History rarely evolves linearly. Technologies often change character unexpectedly.
Deeper issue not economics but power
AI will create enormous wealth and productivity. But it will also create unprecedented concentration of power.
A single engineer equipped with AI agents may one day perform work previously requiring entire departments. Firms with superior AI infrastructure may scale globally with remarkably few workers. Some nations may become vastly more productive than others.
But the same process renders many forms of labour economically redundant. The divide between capital owners and labour, AI-enabled firms and traditional firms, technologically dominant nations and dependent nations and high-skill and low-skill workers may widen dramatically. From productivity revolution, AI may evolve into a concentration-of-power revolution.
AI may become humanity’s most powerful dual-use technology
The deeper concern is that AI possesses the potential to breach ethical, societal and even civilisational boundaries at enormous scale.
Like nuclear technology, AI is fundamentally dual use. The same systems that can accelerate scientific discovery, optimise industries, improve healthcare and raise productivity can also manipulate populations, automate cyber warfare, destabilise political systems, enable autonomous weapons, amplify misinformation, intensify surveillance and erode trust itself.
Unlike nuclear technology, however, AI is far easier to distribute, replicate and embed invisibly into society. Its risk may therefore become more pervasive and harder to contain.
AI’s physical demands are also often underestimated. It requires semiconductors, data centres, electricity grids, cooling systems, fibre networks, strategic minerals and enormous water consumption.
As AI scales globally, it will increasingly compete for energy, minerals and water resources. AI therefore reshapes geopolitics not merely digitally, but physically.
The most uncomfortable question may be human
What happens to human discipline, curiosity and cognitive development when intelligence itself becomes abundant and instantly available?
If generative AI, agentic AI and eventually general AI can write, analyse, code, reason, negotiate, create and strategise, what incentives remain for humans to struggle through the slow and often painful process of learning?
Human capability historically emerged through friction: memorisation built cognition, writing sharpened reasoning, repetition built mastery and problem-solving built resilience.
But if AI increasingly removes friction from thinking itself, humanity may risk outsourcing not merely labour but also cognitive development. Convenience may come at the cost of depth.
The danger is not merely that AI becomes more intelligent than humans. The greater danger may be that humans gradually stop exercising their own intelligence because AI becomes too useful, too efficient and too convenient to resist.
The likely outcome: Split dominance
The most probable outcome may not be total domination by either side. Instead, the world may split across layers of AI stack.
America continues dominating frontier foundation models, AI operating systems, AI chips, global software ecosystems, scientific AI breakthroughs, high-margin AI monetisation and capital-market financialisation.
China dominates AI deployment, AI manufacturing ecosystems, robotics at scale, cost-efficient AI products, AI-enabled infrastructure, industrial AI diffusion and physical economy integration.
In other words, America may monetise AI and China may industrialise AI. This is consistent with the broader argument of our earlier article:
• America compounds financial and technological dominance.
• China compounds industrial and ecosystem dominance.
Together, they may dominate different layers of the global AI systems.
Conflict may not be inevitable
Many modern geopolitical analyses continue interpreting US-China relations primarily through the Western lens of inevitable strategic conflict — where rising powers must eventually confront existing power.
But many Asian political traditions historically evolved less around absolute ideological confrontation and more around coexistence, layered influence, pragmatic interdependence and commercial exchange despite rivalry.
Imperial China’s tributary system, for example, often functioned not merely as territorial domination, but simultaneously as hierarchy, trade architecture and political stabilisation. Even today, China’s Belt and Road strategy reflects not just geopolitical expansion, but also an attempt to deepen economic interdependence, industrial connectivity and long-term commercial integration.
This does not imply the absence of competition. Great powers will still compete intensely on technology, capital, standards, energy, security and influence. But competition does not necessarily require total decoupling or civilisational conflict, or war.
In fact, the rational outcome for both America and China is not mutual destruction, but mutual dependence large enough that both continue benefiting while competing aggressively. The deeper question may therefore not be whether America and China can coexist. It may instead be whether the rest of the world becomes structurally dependent on whichever ecosystem compounds faster.
Now to Asean
China’s industrial expansion carries profound implications for Asean economies.
First, Asean will likely receive significantly more Chinese investments: factories, logistics hubs, industrial parks, and AI-enabled manufacturing and regional distribution networks. But investment alone does not necessarily create full industrial upgrading.
The danger is that higher-value technology, machinery, engineering, software, component ecosystems, strategic supply chain and intellectual property remain China-centred. Asean risks becoming an assembly base, labour pool, land provider, tax-incentive platform and consumption market.
Second, local firms will face enormous competitive pressure. Chinese firms bring scale, low costs, rapid product cycles, integrated supply chains and increasingly sophisticated technology. Many Asean manufacturers may struggle to compete directly.
Third, consumers benefit first but domestic industry may weaken later. Cheaper Chinese products improve living standards initially. But if local productive capacity weakens over time, labour income, industrial capability and technological depth may eventually suffer.
This is why the real question for Asean is no longer “How much foreign investment came in?” but increasingly “What capability remained behind?”
Did the investment create local suppliers, engineering capability, domestic ownership, technology transfer, export competitiveness, regional brands and an indigenous ecosystem? Or merely low-value assembly work (for foreign workers, in the case of Malaysia)?
Asean still retains agency
Asean should not view itself merely asa passive recipient of industrial gravity. Countries such as Singapore, Vietnam, Indonesia and Malaysia still possess significant strategic agency.
The region can still negotiate technology transfer, build sovereign data and AI capabilities, strengthen domestic capital formation, coordinate industrial policies selectively, deepen regional supply chains and develop specialised industrial niches.
Asean cannot realistically replicate China across everything. China’s scale is too large, its ecosystem depth too extensive.
But Asean countries can still build specialised depth in areas such as semiconductor packaging, energy infrastructure, commodity processing, regional logistics, data centres, AI services, robotics maintenance, industrial software, healthcare manufacturing and cross-border digital infra. The challenge is not whether Asean competes with China directly but whether Asean merely becomes economically dependent on China’s ecosystems or uses proximity to them to build its own enduring ecosystem.
Historically, Asean has rarely functioned as a truly unified economic bloc. That weakens its negotiating leverage.
The real transformation
Ultimately, AI will not just reshape economies and industries. It may reshape the balance between:
• Labour and capital;
• States and corporations;
• Humans and machines;
• Convenience and discipline;
• Intelligence and wisdom itself.
And history suggests that technologies powerful enough to transform civilisation are also powerful enough to destabilise it.
