For Southeast Asia, the AI Race Is About Access, Not Supremacy

The global artificial intelligence (AI) debate is dominated by frontier models. Each new release from American and Chinese firms is compared according to benchmark performance, reasoning ability, context length and progress towards artificial general intelligence. This framing presents the AI race as a competition to build the smartest model.

Recent visits and investment announcements by American and Chinese technology firms have reinforced the impression that Southeast Asia is becoming an important arena of AI competition. Yet for most governments, firms, universities and citizens in the region, frontier supremacy may not be the most important contest. The key is whether they are available and at an appropriate price point.

The practical value of AI depends on whether people can afford to use it, whether it works in local languages, whether the infrastructure supporting it remains available and whether foreign governments or companies can change the conditions of access. A theoretically superior model may have limited value if it is too expensive, performs poorly in regional languages such as Bahasa Indonesia or Vietnamese, or becomes unavailable because of a commercial decision, cloud disruption or export restriction.

For Southeast Asia, the AI race is therefore also a race over the price and availability of useful intelligence.

AI is often treated as a software product accessed through a browser. In reality, it depends on a long and capital-intensive stack: advanced semiconductors, cloud platforms, data centres, electricity, cooling systems, fibre networks, subsea cables, foundation models, development tools and technical labour.

Control over these layers produces both commercial and geopolitical power. Companies that control these stacks can bundle services, attract developers and make customers dependent on their ecosystems. Governments that influence these companies can also shape which countries, firms, and users receive access to advanced AI models and under what conditions.

This means that Southeast Asian dependence does not begin or end with the choice of a chatbot. Dependence becomes a vulnerability when a supplier cannot be replaced quickly, affordably and without serious disruption.

In the AI race, Southeast Asia’s room for manoeuvre will depend less on who builds the smartest model than on whether the region retains meaningful choices over which technologies it can access, how they are deployed, and on what terms.

The US and China are both trying to develop powerful models, advanced chips, major technology firms and influence over global AI standards. The main difference lies in the constraints under which their companies operate.

American frontier laboratories benefit from relatively abundant access to advanced chips, hyperscale cloud infrastructure, deep capital markets and large enterprise customers. These conditions encourage scale-intensive development, vertical integration and proprietary platform strategies. For users, this pathway can offer strong frontier performance, but it can also tie them to proprietary clouds, security systems and regulatory decisions made in the US.

Chinese firms face tighter access to advanced computing hardware because of US export controls. These constraints have given Chinese firms stronger reasons to improve efficiency, adapt models to less advanced or domestically produced chips, reduce prices and release open-weight models that others can modify and deploy. However, Chinese platforms raise concerns about data governance, censorship, cybersecurity, and political influence.

These are differences of degree rather than two sealed national models: American firms also pursue efficiency and open-weight releases, while Chinese firms continue to invest in scale and frontier performance. For Southeast Asia, the relevant question is not which national ecosystem is universally superior, but which combination of technologies provides acceptable capability, cost, openness, governance and reliability for a particular task.

Southeast Asia’s AI access problem has at least four dimensions.

The first is price. As AI firms attempt to recover the costs of large infrastructure investments, a two-tier market may emerge: frontier intelligence for users able to pay premium prices and “good enough” intelligence for everyone else. For many regional public agencies and SMEs, price differences matter more than marginal differences in benchmark performance.

The second dimension is political availability. Access to frontier systems may increasingly be shaped by national security policy. Countries that are neither adversaries nor priority domestic markets may face delayed releases or restricted features.

Third, access is linguistic. A cheap model that performs well in English but poorly in Thai, Khmer, Lao, Burmese, Tagalog, Malay, or Tamil may be affordable but still ineffective. In healthcare or public administration, local-language performance is highly consequential.

The fourth dimension is switching risk. AI systems become embedded in data pipelines and organisational routines. A cheap pilot may become costly to scale, or a public agency may discover that transferring its system to another provider is operationally prohibitive.

Therefore, great-power competition in the AI frontier is a double-edged sword: it may deliver cheaper and more capable AI to Southeast Asia, but such competition can also create new bottlenecks, political conditions, and infrastructure risks.

As in other areas of great-power competition, many Southeast Asian governments are likely to hedge rather than align exclusively with one technological ecosystem. The region is unlikely to reproduce the entire AI value chain, nor would attempting to do so necessarily be a sensible use of public resources. Instead, Southeast Asian governments and companies are already beginning to combine technologies across ecosystems rather than adopting a single national stack wholesale.

Policy Implications

The strategic challenge for the region is ensuring that these combinations expand real options rather than merely create more complicated forms of dependence. Technological hedging should therefore mean more than purchasing from multiple suppliers. It requires preserving credible alternatives across the stacks and maintaining the practical ability to move data, applications, and workloads when prices, suppliers or access conditions change.

The goal is not technological autonomy, which is unrealistic for most Southeast Asian states. The objective should be resilient access: ensuring that essential AI remains affordable, locally effective, continuously available, and replaceable when suppliers or political conditions change. In the AI race, Southeast Asia’s room for manoeuvre will depend less on who builds the smartest model than on whether the region retains meaningful choices over which technologies it can access, how they are deployed, and on what terms.

2026/253

---

This is an adapted version of Trends in Southeast Asia 2026/24, published on 14 September 2026. The paper and its references can be accessed at this link.