Southeast Asia's digital economy set for $368B GMV in 2026, tracking to $1T by 2035

For a small electronics retailer in Jakarta or a logistics operator in Ho Chi Minh City, the past 36 months have rewritten the playbook. Generative AI tools that once required specialized expertise now sit inside the same smartphones used to scan inventory, draft supplier emails, and translate customer chats. Across Southeast Asia, that shift has been unusually compressed: 243 million people—58% of the region's working population—have adopted generative AI in under three years, scaling four times faster than mobile internet did in its early phase. Yet the same retailers and operators now confront a harder question: how to convert that breadth of adoption into measurable economic output before the next investment cycle locks in.

The Macro Picture: $368 Billion and a Token Gap

The 11th annual e-Conomy SEA report, released today by Google, Temasek, and Bain & Company, sizes Southeast Asia's digital economy at $368 billion in gross merchandise value for 2026, up 17% year-on-year, with a trajectory that puts the region past $1 trillion by 2035. To capture the depth of AI activity, the report introduces a new unit of measurement—tokens processed per capita—positioning it as the AI equivalent of GMV. Current regional use sits at 5 to 7 quadrillion tokens per capita, roughly 10% of US levels. Consumers account for about 60% of those tokens, mostly for everyday work and study tasks, while enterprise workloads remain comparatively thin, except in Singapore and Malaysia, where they drive 45% to 60% of usage.

That asymmetry is the central tension of the report. Adoption has been near-vertical; depth has lagged. As users move from single-turn queries to delegating multi-step tasks to AI agents, regional token demand is projected to scale by up to 60-fold by 2030. Whether that demand materializes as productive output—or as a wave of underutilized subscriptions—will depend on infrastructure, governance, and the willingness of enterprises to redesign workflows rather than simply bolt AI onto legacy processes.

Infrastructure, Energy, and the Cost of Catching Up

The operational friction is most visible in compute. To meet projected demand, the region needs roughly three times more data center capacity over the next five years, with AI workloads claiming a larger share of total infrastructure. Private capital is responding: deal value in AI compute and digital infrastructure reached $20 billion over the last 12 months, anchored by the $11 billion acquisition of ST Telemedia Global Data Centres. Yet capital alone does not solve the energy equation. Singapore, the region's primary hub, is allocating an additional 200 megawatts of green energy to its 1.4-gigawatt compute base and binding itself to Power Usage Effectiveness requirements under the Digital Infrastructure Bill. Malaysia is advancing a parallel model through its Corporate Renewable Energy Supply Scheme.

Inside enterprises, the friction is more granular. E-commerce, the region's largest sector at $226 billion in GMV and 18% year-on-year growth, is already showing what embedded AI can deliver: a 10% conversion lift and up to 75% higher average order value among early adopters. Online travel ($59 billion, +10%) has seen standalone GenAI platforms overtake social media as the secondary planning tool, capturing 26% of journeys versus social's 23%. Online media ($44 billion in revenue, +22%) is the fastest-growing sector, with AI-driven advertising delivering 10% to 20% engagement uplifts without proportional budget increases. Mobility and food delivery ($40 billion, +15%) report up to a 28% surge in ride-hailing tied to predictive home-screen suggestions and route optimization. Digital financial services, the largest pool by gross transaction value at $1,650 billion and +16% year-on-year, is applying AI across credit assessment, fraud detection, and customer acquisition, with digital lending balances up 22% to $116 billion. The pattern across sectors is consistent: AI is moving from back-office optimization to front-end product experience, and the enterprises that redesign workflows around it are extracting measurable returns. Those that treat AI as a feature add-on are not.

Voices from the Ground

The report's authors frame the moment as a transition from breadth to depth, and they are explicit about the conditions required to sustain it.

Southeast Asia is entering an AI decade defined by rapid adoption, economic opportunity, and a distinct pragmatic optimism toward AI. As people move from basic queries to trusting AI with more complex tasks, Southeast Asia has a unique window to lead the world in agentic commerce. Unlocking that full potential will require clear governance, transparent capabilities, and an unwavering commitment to responsible AI deployment across the whole ecosystem.

Southeast Asia's AI story is shifting from how many people use it to how deeply they use it. Adoption is already among the fastest the region has seen, yet usage per person is a fraction of US levels, and most users remain on free tiers. Closing that gap will drive the next wave of growth. As token costs fall and enterprise workloads scale, AI capabilities will become increasingly accessible. What will differentiate them is how fast they embed it across the customer journey, from discovery to conversion.

Consumer sentiment data in the report reinforces that framing. Seventy-two percent of users already rely on everyday automations; 75% express willingness to let AI agents autonomously find the best prices and complete purchases on their behalf. Sixty-two percent report greater trust in AI tools than a year ago, driven by proven track records (56%), growing familiarity (55%), and output accuracy (53%). That trust, however, is conditional: 44% demand robust privacy protections, and 30% require the ability to undo autonomous actions.

Strategic Outlook: From Adoption to Embedding

The next 12 to 24 months will test whether Southeast Asia can convert its adoption lead into a productivity lead. Four conditions identified in the report will determine the outcome: accelerated agentic adoption backed by model reliability and privacy safeguards; scaled AI infrastructure with a roughly threefold expansion in data center capacity; model and energy efficiency anchored by national frameworks such as Singapore's binding PUE requirements and Malaysia's CRESS model; and sustainable economics, meaning falling inference costs paired with demonstrable ROI.

Singapore is positioned as the region's proving ground. Its digital economy is projected to reach $31 billion in GMV in 2026, up 8% year-on-year, and $59 billion by 2035. GenAI penetration has reached 70% of the working population, or 3.2 million people, and per capita AI use is already nearly half the US level and about twice Europe's, by the report's estimate. The country's Model AI Governance Framework for Agentic AI and the Safeguards for Agentic Finance at Runtime (SAFR) provide the regulatory scaffolding; the National AI Impact Programme aims to train 100,000 individuals by 2029.

Capital flows are following the same concentration. Companies with core AI products captured 80% of deal value in the last six months, up from 30% in the second half of 2024. Private funding for the region's digital applications sector has stabilized at approximately $8 billion annually, with AI-related investments claiming a growing share. The upcoming ASEAN Digital Economy Framework Agreement, expected in November 2026, will establish the world's first region-wide digital trade pact. For enterprises operating across the six largest Southeast Asian markets—Indonesia, Malaysia, the Philippines, Singapore, Thailand, and Vietnam—the agreement offers a framework for cross-border data flows, digital payments, and AI governance that has so far been assembled piecemeal.

The strategic playbook for enterprise leadership is therefore less about whether to adopt AI and more about where to embed it. The report's data suggests that the highest returns are accruing to organizations that treat AI as a workflow redesign problem rather than a tooling decision—integrating it into discovery, comparison, conversion, and post-purchase service rather than layering it onto existing interfaces. As token costs continue to fall and enterprise workloads scale, the differentiator will be execution speed against that workflow thesis, not the speed of any single deployment.