World Bank advocates for Thailand to adopt a pragmatic AI strategy over an arms race

A new World Bank report urges Thailand to focus on localised AI tools and incremental advancements instead of competing in the costly frontier AI race, highlighting a practical approach centred on existing strengths and specialised applications.

Thailand should not try to win an arms race to build the largest artificial intelligence models, according to a new World Bank assessment that argues the country would gain more by adopting existing tools, adapting them to local conditions and advancing into more specialised systems when ready. The report says the cost of frontier AI development is now so high that even major technology groups can outspend many national economies, making a first-principles race unrealistic for most developing countries. It instead recommends a practical path centred on Thai-language tools, local data and narrowly targeted applications that can run on ordinary devices or work offline.

The World Bank frames that strategy as “Adopt, Adapt and Advance”. In its view, countries should begin by using ready-made systems, then tailor them to local laws, culture and working practices before eventually building more advanced capabilities of their own. The report contrasts that approach with the scale of investment now being poured into frontier AI by Alphabet, Amazon, Meta, Microsoft and Oracle, and argues that Thailand would be better served by focusing on productivity gains than on trying to replicate the full technology stack from scratch. That means developing “Small AI” for specific tasks rather than chasing universal models that demand vast computing power, data centres and research budgets.

The report also suggests that Thailand already has a role in the AI economy. It ranks the country among the top exporters in the developing world for goods that support AI systems and places it among the 10 largest recipients globally for data-centre investment. The findings point to a manufacturing base that includes electronics, semiconductors and related infrastructure, but the World Bank says that hardware alone will not capture the greatest share of value. To move up the chain, Thailand would need stronger capabilities in design, software, cloud services, data services and higher-value applications.

Data on trade and investment show how quickly the field is expanding. The World Bank says goods linked to AI development increased their share of global trade from about 13% in 2023 to nearly 17% by the end of 2025, with low- and middle-income countries accounting for roughly a third of export value under one definition and about 39% under a narrower one. China, Mexico, Malaysia, Vietnam and Thailand were the leading developing-country exporters in 2025. At the same time, data centres accounted for more than one-fifth of global greenfield foreign direct investment in 2025, although the report notes that their job creation and environmental impact remain contested.

For Thailand, the labour market implications are more complex than simple job losses. The World Bank estimates that about 4.5% of jobs in low- and middle-income countries involve tasks that generative AI could replace, while 16.2% are more likely to benefit from AI as a productivity tool. The study says the earliest exposure is likely in business services, call centres, office support, finance and software, while many jobs in agriculture, retail, hotels and small businesses still rely heavily on physical work and direct customer contact. The report also cautions that these are regional averages, not a direct forecast for Thailand.

Businesses are already experimenting. The World Bank surveyed firms in India, Jordan, Kenya, Mexico, Nigeria and Thailand, including 360 Thai companies, and found that around one in five small businesses were using AI chatbots, compared with about one in four in the United States. Across the participating countries, entrepreneurs expected AI to lift productivity by about 11% over the next three years, with staff numbers still projected to rise by around 7%, suggesting that many firms see AI as a way to expand output rather than cut jobs.

Public services could also benefit, but only if governments tackle basic constraints first. The report cites examples from India, Kenya and Bangladesh where AI has helped farmers, mediators and doctors, yet says many public agencies struggle with fragmented data, outdated systems, weak internet access and limited staff skills. In Thailand, that means agencies would need Thai-language data in usable formats, stronger oversight of results and processes for maintaining and improving systems over time, rather than simply buying software and expecting it to work.

The World Bank says governments must act as enabler, user and regulator at the same time. That implies investment in electricity, broadband, education, digital skills and computing capacity, along with procurement rules that favour ongoing testing and improvement rather than one-off purchases. It also means reviewing existing laws, using industry standards as a starting point and managing risks tied to privacy, discrimination, safety, fraud and cybercrime. For Thailand, the report’s core message is clear: the opportunity lies not in competing with the world’s biggest AI builders, but in turning existing strengths into useful, localised technology that lifts productivity across the economy.