Asia Shifts from Digital Transformation to Global Leadership

For most of the past decade, “digital transformation” in Asia meant catching up: banks digitising branches, retailers building e-commerce rails, governments rolling out digital ID. That era is closing. The region is no longer principally absorbing technology built elsewhere. It is building the foundational layers of the AI economy itself, exporting the compute, the chips and, increasingly, the models that the rest of the world depends on.

The shift shows up first in trade data. Asian economies contributed close to two-thirds of the growth in global AI-related trade in 2025, and semiconductor sales are on track to approach the $1 trillion mark globally in 2026, with the Semiconductor Industry Association crediting AI infrastructure spending as the primary driver. Asia is not simply plugged into that boom. It is the boom’s supply base, and its capital.

The semiconductor foundation

The starting point for Asia’s claim to technological leadership remains hardware. Including Japan, the region now accounts for roughly 72 percent of global semiconductor production and about 95 percent of the most advanced chips used in AI accelerators, according to analysis from State Street Global Advisors. Taiwan manufactures the leading-edge logic chips that power frontier AI models, South Korea dominates high-bandwidth memory, and Japan supplies the specialty chemicals and precision equipment that make the rest of the chain possible. J.P. Morgan Asset Management has gone further, putting Asia’s share of leading-edge semiconductor production above 95 percent and noting that technology now makes up more than 30 percent of the MSCI Asia Pacific ex Japan Index.

That concentration is not incidental. It took roughly fifteen years and tens of billions of dollars of sustained investment to build, and it now anchors what has been described as Asia’s industrial supercycle, in which AI data centres, energy infrastructure, defence electronics and supply chain diversification reinforce one another rather than functioning as separate trends. Every advanced GPU shipped anywhere in the world still traces back to Taiwanese silicon, Korean memory and Japanese equipment. The West’s hyperscalers are spending unprecedented sums on AI infrastructure, but a large share of every dollar spent ultimately flows back to Asian suppliers.

Sovereign AI, not borrowed AI

What has changed more recently is the model layer. Through 2024 and into 2025, most of Asia’s AI activity was about deploying Western large language models. That is no longer the whole story. Every major Asia-Pacific economy now runs its own domestic large language model programme: India’s Sarvam AI, Malaysia’s ILMU, Indonesia’s Sahabat AI, Singapore’s SEA-Lion, South Korea’s HyperCLOVA X Think, Japan’s NTT Sarashina and Taiwan’s TAIDE, according to a market report from Digital in Asia. More than 100 countries signed the Bangkok Declaration in February 2026 committing to AI sovereignty, treating dependence on a foreign AI supplier as a strategic vulnerability comparable to energy dependence.

The scale of the underlying commitment is considerable. China’s AI capital expenditure reached an estimated $91 billion to $98 billion in 2025, up more than 50 percent year on year, while its science and technology budget hit roughly $171 billion, the largest single science allocation in the country’s budgetary history, per the same report. East Asian institutions now hold more than 80 percent of global AI patent grants and close to 35 percent of global AI research output. The region that spent the 2010s licensing technology from Silicon Valley is now filing the patents.

Regionally, ASEAN has formalised the same instinct through policy. The sixth ASEAN Digital Ministers’ Meeting, held in Hanoi in January 2026, adopted the Hanoi Digital Declaration and set out the new ASEAN Digital Masterplan 2026-2030, covering AI cooperation, resilient digital infrastructure, a future-ready digital workforce and trusted cross-border data flows. The framing captures the shift in a single phrase: from connectivity to connected intelligence.

The compute build-out

Sovereign models need sovereign compute, and Southeast Asia’s data centre buildout has moved from consumption to infrastructure formation. Greenfield investment in ASEAN’s information and communication sector rose 43 percent to reach $30 billion, driven by demand for data centres, cloud infrastructure and data processing, while investment in electronics and electrical equipment climbed 15 percent to $31 billion. Amazon Web Services, Google and Microsoft have all committed multi-billion-dollar cloud region investments across Thailand, Malaysia and Indonesia.

Thailand has become one of the more visible beneficiaries of that reallocation. The Board of Investment approved a $25 billion data infrastructure commitment from a TikTok affiliate, part of a broader $29 billion package spanning six major projects, with AI processing and cloud services now accounting for 86 percent of total digital investment value approved. Gorilla Technology has separately moved to build a 200 megawatt AI compute campus in Korat, designed to serve Southeast Asian demand for sovereign compute capacity from governments, hyperscalers and enterprises alike. A further $1.99 billion in BOI-approved investment in July, led by AI infrastructure and advanced electronics projects, reflected a 2.4-fold year-on-year surge in investment applications.

Where Thailand fits

Thailand’s position in this shift is more structural than opportunistic. The country is on track for record exports this year, driven substantially by AI-related electronics demand, with the Thai stock market emerging as Southeast Asia’s strongest performer this year and Delta Electronics alone gaining roughly 80 percent as investors price in the country’s role in the AI hardware supply chain. The government’s FastPass Program has unlocked an estimated $21 billion in strategic high-tech investment, and Thailand’s AGIBOT APC 2026 initiative signals ambitions to become a regional hub for embodied AI and advanced robotics deployment, not merely a manufacturing site for someone else’s hardware.

That ambition extends to robotics more broadly. Vietnam, Thailand and Malaysia are betting on automation to make their factories attractive for reshoring and near-shoring investment as global supply chains diversify away from China, competing on integration sophistication rather than labour costs alone. Singapore has taken the commercialisation lead within that trend, anchored by a $125 million funding round for 3D-mapping startup dConstruct and a state-backed accelerator that has drawn partners including Nvidia, OpenAI, Grab and DHL.

The caveat

None of this converts automatically into durable leadership. The World Economic Forum, drawing on BCG research, has been explicit that Asia-Pacific’s structural advantages in chips, memory and manufacturing will not translate into AI leadership unless organisations treat AI as a leadership question rather than a technology function delegated downward. The same report notes the window for converting potential into leadership is real but not permanent.

Talent remains the binding constraint. China faces an estimated shortfall of five million AI professionals, Japan a gap of around 100,000, and India reports an 82 percent talent shortage rate for AI-specific roles, even as the region produces AI graduates at record scale. An estimated 87 percent of Chinese-educated AI researchers who move to the United States for work stay there, a brain drain that undercuts the sovereignty ambitions driving so much of the region’s policy activity.

Regulatory fragmentation carries a similar risk in the commercial layer. Southeast Asia’s fintech sector illustrates the pattern: the region has consolidated its position as the world’s most dynamic fintech market, yet a company that reaches product-market fit in Thailand still confronts a materially different regulatory regime in Indonesia, Vietnam or the Philippines, which slows the kind of regional scaling that a genuine leadership position would require. The same tension applies to AI: eleven jurisdictions building sovereign models and sovereign compute simultaneously is a hedge against dependency, but it is not yet the coordinated bloc that would let Asia set global standards rather than simply supply global infrastructure.

The direction of travel, even so, is unmistakable. Asia spent the last technology cycle adopting what others built. It is entering this one building what others will need to adopt.