The largest pool of AI money in Southeast Asia has published where it is going for the next five years, and almost none of it is going to startups.
Temasek closed its financial year in March at a net portfolio value of S$518 billion, up 10.5 percent in Singapore dollars and 14.8 in American ones. AI-related exposure sits at roughly six percent of that. The plan takes it to between ten and fifteen percent by March 2031, moving about S$31 billion of exposure toward something between S$52 and S$78 billion.
Then read the five segments it named: energy and data centres, semiconductors, cloud service providers, foundation models, AI applications and software infrastructure.
Four of those five are things a company in Jakarta cannot become.
Four of those five are things a company in Jakarta cannot become.
They are capital-intensive, land-hungry, licence-bound businesses priced on contracted capacity and cost per megawatt. A Series A software company competes for that money the way a restaurant competes for a port concession.
The regional tables already show the result. Ninety-eight equity deals across Southeast Asia in the first quarter of 2026, the lowest count in at least eight years, with a single data centre operator’s round taking most of the value. Indonesia, once the largest startup market in the region, recorded five deals.
Seed funding for the entire region came to 105 million dollars for the quarter.
That is the whole seed pool, for seven hundred million people, in a boom.
So the AI capital surge is real and it is an infrastructure surge, and a founder reading the headline as a signal about their own raise is reading someone else’s balance sheet.
The useful part of the announcement was the sentence that came after the allocation.
Dilhan Pillay, Temasek’s chief executive, put the other side on the record at the July briefing: the rubber hits the road in AI adoption, and the remaining 85 percent of the portfolio must be focused on AI adoption for competitiveness.
The 85 percent is a purchase order wearing the clothes of a strategy update.
The 85 percent is a purchase order wearing the clothes of a strategy update.
It is the banks, ports, hospitals, telcos and manufacturers Temasek already owns, and their largest shareholder has just told them to adopt or lose ground. They want something working by the next board meeting, and they will pay for it out of an operating line, on a timetable no investment committee moves at.
That budget is larger than the region’s venture pool, and it sits closer to a founder than the venture pool does.
So the practical move for anyone with a term sheet conversation this quarter is to work out, before the second meeting, which side of the split the fund is standing on.
A fund building the fifteen percent asks about contracted capacity, cost per megawatt, and a ten-year payback. A fund serving the eighty-five asks which incumbent is already paying, how much, and whether the contract renews. Answering the first set of questions with software metrics costs a quarter.
The second conversation is the raisable one in this market, and it turns on a signed customer instead of a projection.
Which inverts the usual sequence. The regional playbook has been to close the round and then go hunting for the enterprise logo; the arithmetic above says get the logo first, because it has become the cheaper of the two and the one that makes the other possible.
Temasek will spend up to S$78 billion proving that AI matters. The companies it already owns have to prove it works by Friday.1
Footnotes
-
The counter-case: enterprise procurement in this region is slow, and a purchase order from a Temasek-linked incumbent can take three quarters to convert. That is fair, and it is still faster than a Series A round the fund’s mandate was never going to approve. ↩