Every few months, Southeast Asia’s startup ecosystem gets a headline that sounds like a resurrection. This October brought two. Tracxn reported that tech funding in the region rose 176 per cent to US$13.5 billion in the first nine months of 2026. Dealroom and Founders Forum went further, saying funding had tripled and that Singapore had swallowed 91 per cent of it.
Cue the champagne. But before anyone pops it, look at what happened in the US this week, because it explains what is really happening in Johor.
Big Tech has run out of its own money
On Monday, Meta said it plans to raise as much as US$95 billion by selling new shares. About US$35 billion will come through underwritten offerings. Another US$60 billion will come through an at-the-market programme of newly created Class C shares that carry no votes.
Also Read: Anthropic, DeepSeek and drones: 10 takeaways from the 2026 unicorn boom
This is a company that just reported US$64.6 billion in quarterly revenue. It still expects to spend between US$135 billion and US$145 billion on capital expenditure this year. Its advertising machine, lucrative as it is, no longer covers the bill on its own.
The same day, SpaceX disclosed an agreement involving roughly US$40 billion in principal payments to lease AI chips and data centre hardware from an entity managed by Apollo Global Management. Not buy. Lease. Meanwhile, OpenAI is reportedly raising up to US$30 billion at a US$1.25 trillion pre-money valuation, with Coatue and UAE investors at the table.
Strip away the jargon and the message is simple: the AI build-out has outgrown cash flow. It is now financed by dilution, structured credit and Gulf money. That is not a scandal. It is, however, a change in the risk profile of the whole industry. And Southeast Asia has quietly plugged itself into the bottom of that pipeline.
Three cheques do not make an ecosystem
Read the Tracxn numbers past the headline. The number of funding rounds actually fell 28 per cent, to 189. Three deals, two by Singapore-headquartered data centre operator DayOne and one by video-generation firm Kling AI, added up to US$7.3 billion.
That is 54 per cent of the nine-month total and 85 per cent of the year-on-year increase. Late-stage money more than tripled to US$10.2 billion. Series A and B rounds, the stage where tomorrow’s Grabs are supposed to be built, drew US$3 billion across 80 deals.
Also Read: The AI wrapper reckoning has reached SEA’s funding tables
So the “rebound” is mostly a story about concrete, chillers and power substations. That is useful infrastructure. But it is not a story about venture appetite for Southeast Asian founders, and we should stop narrating it as one.
The equity is the small part
Here is where the costume slips. DayOne’s equity rounds are what show up in startup funding trackers. The debt does not.
Earlier this month, a consortium of seven banks, including CIMB, Maybank, DBS and Standard Chartered, added RM15 billion (about US$3.67 billion) to DayOne’s Johor financing. That takes its debt platform there to RM30 billion, or roughly US$7.3 billion. The company has also reportedly held early talks on a US$500 million bond, and it filed for a Nasdaq listing on 5 October.
That filing is the most honest document in this whole saga. It shows Malaysia generated 87 per cent of DayOne’s first-half revenue, and that ten customers in Johor account for about 1.4 gigawatts of bookings. Ten. Those tenants are, by and large, the same hyperscalers and AI labs now funding themselves with voteless shares and leased chips.
The chain runs like this. Southeast Asian banks lend to a Singapore operator. That operator builds in Johor for a handful of foreign tenants. Those tenants’ AI spending is increasingly funded by capital markets betting on returns nobody has yet proven. If any link wobbles, the wobble travels downstream. Johor is downstream.
“But these are real assets”
The defenders of the build-out have a fair point, so let me make it for them. Data centres are not WeWork. They are physical assets, typically leased on long contracts to customers with enormous balance sheets. A Crédit Agricole CIB banker quoted in the DayOne deal coverage projected Malaysia’s capacity at 4.2 to 6.1 gigawatts by 2030. Someone will need to host the region’s AI workloads.
Also Read: Southeast Asia’s funding winter is real. Its building winter isn’t
All true. But a long lease is only as reassuring as the tenant’s willingness to keep paying for capacity it may not need if the AI revenue story disappoints. Lenders are already growing choosier about smaller infrastructure players whose business depends on future compute demand.
And the clearest winners of an infrastructure boom are often the suppliers, not the owners. Britain’s Computacenter has just raised its profit forecast for the fourth time this year on AI orders. It gets paid whether or not the data centres it equips ever earn a decent return. Southeast Asia, so far, is mostly the landlord, not the supplier.
What we should do differently
None of this means Johor should stop building. It means the region should stop fooling itself about what the numbers say.
First, ecosystem reports, and yes, journalists like me, should publish funding figures with and without infrastructure. A data centre Series C and a Series A for a Vietnamese logistics startup are not the same animal. Lumping them together hides how thin early-stage funding has become.
Second, regulators in Malaysia and Singapore should ask how concentrated their banks’ data centre exposure is. What does it look like if two or three tenants slow their expansion? When one operator’s financing reaches US$7.3 billion and its revenue rests on ten customers, that is a supervisory question, not just a commercial one.
Third, governments handing out land, power and tax breaks for data centres should demand something back for local builders. That means affordable compute for regional startups, research access and local hiring commitments. Otherwise, Southeast Asia rents out its electricity and water to train other people’s models, while its own founders still cannot afford to run theirs.
Also Read: SEA’s AI boom has a water problem it cannot offset away
The AI boom may yet justify every dollar being borrowed. I hope it does. But when Meta has to sell voteless shares and SpaceX has to lease its chips, the smartest thing Southeast Asia can do is admit that its “recovery” is riding the same wave. Then it should start asking who is holding the surfboard when that wave breaks.
The post Meta sells voteless shares, SpaceX leases its chips. Johor should be paying attention appeared first on e27.
Every few months, Southeast Asia’s startup ecosystem gets a headline that sounds like a resurrection. This October brought two. Tracxn reported that tech funding in the region rose 176 per cent to US$13.5 billion in the first nine months of 2026. Dealroom and Founders Forum went further, saying funding had tripled and that Singapore had swallowed 91 per cent of it.
Cue the champagne. But before anyone pops it, look at what happened in the US this week, because it explains what is really happening in Johor.
Big Tech has run out of its own money
On Monday, Meta said it plans to raise as much as US$95 billion by selling new shares. About US$35 billion will come through underwritten offerings. Another US$60 billion will come through an at-the-market programme of newly created Class C shares that carry no votes.
Also Read: Anthropic, DeepSeek and drones: 10 takeaways from the 2026 unicorn boom
This is a company that just reported US$64.6 billion in quarterly revenue. It still expects to spend between US$135 billion and US$145 billion on capital expenditure this year. Its advertising machine, lucrative as it is, no longer covers the bill on its own.
The same day, SpaceX disclosed an agreement involving roughly US$40 billion in principal payments to lease AI chips and data centre hardware from an entity managed by Apollo Global Management. Not buy. Lease. Meanwhile, OpenAI is reportedly raising up to US$30 billion at a US$1.25 trillion pre-money valuation, with Coatue and UAE investors at the table.
Strip away the jargon and the message is simple: the AI build-out has outgrown cash flow. It is now financed by dilution, structured credit and Gulf money. That is not a scandal. It is, however, a change in the risk profile of the whole industry. And Southeast Asia has quietly plugged itself into the bottom of that pipeline.
Three cheques do not make an ecosystem
Read the Tracxn numbers past the headline. The number of funding rounds actually fell 28 per cent, to 189. Three deals, two by Singapore-headquartered data centre operator DayOne and one by video-generation firm Kling AI, added up to US$7.3 billion.
That is 54 per cent of the nine-month total and 85 per cent of the year-on-year increase. Late-stage money more than tripled to US$10.2 billion. Series A and B rounds, the stage where tomorrow’s Grabs are supposed to be built, drew US$3 billion across 80 deals.
Also Read: The AI wrapper reckoning has reached SEA’s funding tables
So the “rebound” is mostly a story about concrete, chillers and power substations. That is useful infrastructure. But it is not a story about venture appetite for Southeast Asian founders, and we should stop narrating it as one.
The equity is the small part
Here is where the costume slips. DayOne’s equity rounds are what show up in startup funding trackers. The debt does not.
Earlier this month, a consortium of seven banks, including CIMB, Maybank, DBS and Standard Chartered, added RM15 billion (about US$3.67 billion) to DayOne’s Johor financing. That takes its debt platform there to RM30 billion, or roughly US$7.3 billion. The company has also reportedly held early talks on a US$500 million bond, and it filed for a Nasdaq listing on 5 October.
That filing is the most honest document in this whole saga. It shows Malaysia generated 87 per cent of DayOne’s first-half revenue, and that ten customers in Johor account for about 1.4 gigawatts of bookings. Ten. Those tenants are, by and large, the same hyperscalers and AI labs now funding themselves with voteless shares and leased chips.
The chain runs like this. Southeast Asian banks lend to a Singapore operator. That operator builds in Johor for a handful of foreign tenants. Those tenants’ AI spending is increasingly funded by capital markets betting on returns nobody has yet proven. If any link wobbles, the wobble travels downstream. Johor is downstream.
“But these are real assets”
The defenders of the build-out have a fair point, so let me make it for them. Data centres are not WeWork. They are physical assets, typically leased on long contracts to customers with enormous balance sheets. A Crédit Agricole CIB banker quoted in the DayOne deal coverage projected Malaysia’s capacity at 4.2 to 6.1 gigawatts by 2030. Someone will need to host the region’s AI workloads.
Also Read: Southeast Asia’s funding winter is real. Its building winter isn’t
All true. But a long lease is only as reassuring as the tenant’s willingness to keep paying for capacity it may not need if the AI revenue story disappoints. Lenders are already growing choosier about smaller infrastructure players whose business depends on future compute demand.
And the clearest winners of an infrastructure boom are often the suppliers, not the owners. Britain’s Computacenter has just raised its profit forecast for the fourth time this year on AI orders. It gets paid whether or not the data centres it equips ever earn a decent return. Southeast Asia, so far, is mostly the landlord, not the supplier.
What we should do differently
None of this means Johor should stop building. It means the region should stop fooling itself about what the numbers say.
First, ecosystem reports, and yes, journalists like me, should publish funding figures with and without infrastructure. A data centre Series C and a Series A for a Vietnamese logistics startup are not the same animal. Lumping them together hides how thin early-stage funding has become.
Second, regulators in Malaysia and Singapore should ask how concentrated their banks’ data centre exposure is. What does it look like if two or three tenants slow their expansion? When one operator’s financing reaches US$7.3 billion and its revenue rests on ten customers, that is a supervisory question, not just a commercial one.
Third, governments handing out land, power and tax breaks for data centres should demand something back for local builders. That means affordable compute for regional startups, research access and local hiring commitments. Otherwise, Southeast Asia rents out its electricity and water to train other people’s models, while its own founders still cannot afford to run theirs.
Also Read: SEA’s AI boom has a water problem it cannot offset away
The AI boom may yet justify every dollar being borrowed. I hope it does. But when Meta has to sell voteless shares and SpaceX has to lease its chips, the smartest thing Southeast Asia can do is admit that its “recovery” is riding the same wave. Then it should start asking who is holding the surfboard when that wave breaks.
The post Meta sells voteless shares, SpaceX leases its chips. Johor should be paying attention appeared first on e27.
投资人评论
聊判断，不堆黑话登录且账号审核通过后可评论。 立即登录