FlowtandBriefAI · Software · ResearchMeet Flowtand

Google-Led e-Conomy SEA Report Projects $368B Digital Economy, $1T by 2035

Announcement

Google · Nitin GajriaManaging DirectorSoutheast Asia and South Asia Frontier Markets

Browse Industry topic page →

The 11th e-Conomy SEA report, published by Google, Temasek and Bain & Company, puts Southeast Asia's 2026 digital economy at $368 billion in GMV, up 17% year-on-year, and projects it exceeding $1 trillion by 2035. The report ties growth to generative AI adoption by 243 million people and forecasts much greater use of AI agents in commerce.

What the report actually says

The 11th e-Conomy SEA report, published by Google with Temasek and Bain & Company, estimates Southeast Asia's digital economy at $368 billion in GMV for 2026, a 17% year-on-year increase, and projects that it will pass $1 trillion by 2035. The publishers frame generative AI as a core driver: they say 243 million people, or 58% of the working population, use it in daily life, and that adoption is scaling four times faster than mobile internet did. Sector figures include e-commerce at $226 billion GMV, online travel at $59 billion, online media revenue at $44 billion and digital financial services GTV at $1,650 billion. On trust, the report says 72% of users rely on everyday automation features and 75% would let AI agents compare prices and complete transactions, while expecting privacy safeguards and reversible transactions. It also lists conditions for sustaining growth, including roughly tripling data center capacity over five years and policies such as Singapore's power efficiency rules and the planned ASEAN digital trade pact.

A test case for an e-commerce product manager

A marketplace product manager in the region could treat the report's consumer-trust findings as a hypothesis rather than a forecast. The concrete task is to introduce a narrowly scoped agent feature, for example one that compares prices across a shortlist of categories and prepares an order for buyer confirmation, then measures behaviour against a control group that uses standard manual browsing. The report's claim that personalised experiences drive higher conversions is the publishers' assertion; a team could try to reproduce that in its own funnel before extending an agent to unsupervised checkout, since the excerpt indicates that willingness to delegate transactions is stated intent, not observed purchasing.

Trial design and failure checks

A team could begin by defining a baseline from existing session data: search-to-cart rate, cart abandonment and time to purchase for the chosen categories. It could then run a limited comparison with the agent feature, tracking completion rates, reversals and support contacts alongside revenue. Success should require improvement on the primary metric without a rise in reversals or complaints; if either rises, the feature is not ready. The report suggests expecting clarity on limits and an ability to undo agent actions, so the trial should include an explicit confirmation step and a visible undo path, with a person reviewing agent-prepared orders before release to new categories. Because this evidence is a short excerpt, the team's own measurements should outrank the published uplift figures.

Trade-offs and when not to proceed

The report's growth outlook depends on infrastructure and policy conditions that individual firms cannot control: it projects a roughly threefold expansion of data center capacity in five years and points to energy efficiency rules and the planned ASEAN digital trade pact. A business weighing agent features should check its own compute and cost assumptions before assuming that serving models stays affordable. Agentic commerce may not suit high-risk purchases, categories where buyers value direct advice, or markets where users lack the privacy assurances the report says they want. Uncertainties include the report's survey composition and the meaning of its token benchmark, neither of which the excerpt details; readers should consult the full report before citing its figures in planning.

Context and limitations

  • This brief draws on a page excerpt; the full report methodology and sample were not reviewed.
  • Figures such as conversion uplifts and the 60-fold usage projection are the publishers' estimates and have not been independently verified.
  • The token-based benchmark is defined only in the source's notes; its comparison rules are not explained in the excerpt.
  • Based on bounded source text, not independent testing. Practical suggestions are editorial proposals.