Southeast Asia's digital economy to hit $368 billion, driven by AI

Google, Temasek and Bain & Company have put a number on the year. According to their report, Southeast Asia's digital economy will reach $368 billion in gross merchandise value (GMV) in 2026, up 17% from $315 billion in 2025. The projection rises to $620 billion by 2030 and $1 trillion by 2035.

What drives the total matters more to us than the total itself. According to the study, 243 million people in the region already use generative AI actively, which is 58% of the working-age population. That adoption is moving four times faster than mobile internet did in its day.

What Google published

This is the eleventh edition of the e-Conomy SEA report, titled 'Building the Intelligent Economy' this year. Google presented it today, October 21, on its official blog, in a post signed by Nitin Gajria, Managing Director for Southeast Asia and South Asia Frontier Markets. By sector, the breakdown looks like this:

1. E-commerce: $226 billion in GMV, up 18% year on year. It is the largest sector and where the report places the biggest AI impact, with a 10% conversion uplift and order volume increases of up to 75%.
2. Digital financial services: $1.65 trillion in gross transaction value (GTV), up 16%.
3. Online media: $44 billion in revenue, up 22%. Commerce media already accounts for $4 billion, and engagement is up 10% to 20%.
4. Online travel: $59 billion, up 10%.
5. Mobility and food delivery: $40 billion, up 15%.

Online media is the fastest growing sector in relative terms. That does not look like a coincidence, since it is also where recommendation systems and content generation feed most directly into the business.

The agentic data point

For those of us building agents, the most interesting part is in the consumer surveys. 75% of respondents are willing to let an AI agent compare prices and complete transactions on their own. 72% already use automations such as price tracking every day, and 62% say they trust AI more than they did a year ago.

Those percentages need careful reading. Being willing to delegate a purchase is not the same as doing it, and a lot of engineering sits between the two: scoped permissions, confirmation before payment, a log of every action and a clear way to undo mistakes. This is exactly the kind of problem that standards like MCP help with. With MCP, a model calls catalogs, payment gateways or inventory systems through tools with an explicit contract, instead of improvising on a website designed for people.

Infrastructure, the bottleneck

The report estimates that the region will need to triple its data center capacity over the next five years to sustain this growth. The figure affects anyone shipping product there, because latency, data residency and cost per inference all depend on that capacity arriving on time and in the right countries. For a team serving real-time agents, it matters whether inference runs from nearby data centers or from another continent.

Who this is useful for

For product teams selling in the region, the report offers a quantified case for prioritizing agentic features in e-commerce, where the volume is concentrated. For agencies and consultancies, the conversion figures are a good reference, although they are aggregates and do not replace your own testing. And for anyone following the market from Europe, it is a reminder that consumer AI adoption does not move at the same pace everywhere.

It is also worth keeping in mind who signs it. Google sells cloud and AI infrastructure, Temasek is a very active investor in the region's tech ecosystem, and projections for 2035 are speculative by definition. That does not invalidate the 2026 data, but it does suggest treating the trillion dollars as a scenario rather than a firm forecast.

At ElephantPink, we think the number to watch is the 75% willing to delegate purchases. Whether that intent turns into real usage will depend less on the model than on the design of the tools, permissions and controls around it.