Behind the US$7.1 Trillion Infrastructure Bill: Plenty of Money for Computing, but No Clear Power Plan
By Yuanyin in Tokyo
On November 10, the Asian Development Bank released a report titled The $7 Trillion Challenge: Building ASEAN for Tomorrow.
The report estimates that Southeast Asia will need US$7.1 trillion in infrastructure investment between 2026 and 2040, equivalent to an average of US$476 billion a year, or 8.7% of regional GDP.
Public investment can cover only about 45% of that requirement, leaving an annual financing gap of roughly US$264 billion.
The structure of that investment is even more revealing.
Power and transport account for 85% of the total, requiring around US$3.6 trillion and US$2.5 trillion, respectively. Digital infrastructure — including broadband networks and data centers — is allocated another US$584 billion.
One Word That Cannot Be Avoided: Electricity
US$7.1 trillion is a striking number.
But the more important question is how that money is divided.
As ADB President Masato Kanda put it, Southeast Asia’s factories need reliable electricity, while businesses need faster and more secure ways to access markets.
Factories need electricity.
So do data centers.
ADB places spending on data centers under “digital infrastructure” and electricity investment under the “power” category. But ultimately, both sets of investment depend on the same power grids.
And in many parts of Southeast Asia, those grids are already under pressure.
Start with the data-center side.
Electricity demand from Southeast Asian data centers could reach 171 to 258 TWh by 2035, compared with only around 20 TWh in 2023.
Meeting that demand would require an additional 53.5 to 82 GW of generation capacity.
An AI server rack can consume 40 to 200 kW, compared with just 5 to 15 kW for a traditional enterprise rack.
Power density has therefore increased by more than an order of magnitude, fundamentally changing what the grid is being asked to support.
Now consider the electricity side.
A significant share of ADB’s US$3.6 trillion power-investment requirement is tied to cross-border interconnection, including the ASEAN Power Grid.
The logic is straightforward.
Resource-rich Laos has hydropower. Singapore has demand. Cross-border transmission can move electricity from one market to another.
In May, ADB also launched a broader Pan-Asia Grid initiative, seeking to extend cross-border power trading beyond ASEAN into the wider Asia-Pacific.
But the construction timetable for power infrastructure is very different from that of data centers.
A World Economic Forum analysis published in November highlighted the mismatch clearly: data centers can typically be built within two to five years, while new generation, transmission infrastructure and grid reinforcement can require five to eight years or longer.
In Thailand’s Eastern Economic Corridor, substations are already approaching capacity. Facilities requiring more than 50 MW of electricity can face grid-connection waits exceeding 18 months.
ADB itself acknowledges another important problem: the shortage of “well-prepared projects” is a major obstacle to attracting private capital.
In other words, the problem is not simply a lack of money.
There are not enough grid projects that are sufficiently developed, approved and technically ready to absorb the available capital.
The money for electricity may exist on paper.
But engineering schedules, cross-border power rules and queues for grid connection are not determined by financing figures alone.
A Visible Crack: Johor
Johor offers one of the clearest examples.
Its operating data-center capacity increased from 94 MW in 2023 to 1,110 MW in 2026, with another 3,088 MW under construction or planned.
Wood Mackenzie estimates that data centers already account for around 24% of Johor’s electricity consumption, a figure that could rise to 40% by 2035.
But Johor’s grid was not designed for growth at this speed.
The vacancy rate for colocation data centers has fallen to just 0.7%, meaning completed capacity is effectively taken almost as soon as it comes online.
The question is where the electricity for the next wave of projects will come from.
Malaysia tightened approvals in February 2026, suspending ordinary new data-center projects and prioritizing AI and frontier-technology facilities.
Thailand has taken an even more direct approach.
In September, it suspended construction on 49 data-center projects and froze approvals for another 117 projects while four subcommittees worked on standards covering electricity, water resources and environmental requirements.
These measures reveal a growing realization across Southeast Asia:
Data-center capital may arrive quickly.
Electricity does not appear simply because an investment contract has been signed.
How much of ADB’s US$584 billion in digital-infrastructure investment can actually be deployed will depend heavily on how much of the US$3.6 trillion power investment is converted into substations, transmission lines and workable grid-connection agreements.
The Real Structural Problem
The most important part of the ADB report is not the headline US$7.1 trillion figure.
It is the structural mismatch the number exposes.
Infrastructure investment is divided into separate categories — “power” and “digital.”
But AI computing increasingly binds those categories together.
The link between them is grid capacity and connection speed.
And that link is not separately priced in either category.
The US$584 billion allocated to digital infrastructure covers data centers, broadband networks and subsea cables.
The US$3.6 trillion allocated to power covers generation, transmission and cross-border interconnection.
They appear as separate lines in an ADB table.
But at substations in Johor, in grid-connection queues in Thailand’s Eastern Economic Corridor, and in Malaysia’s national grid-upgrade plans, they are part of the same problem.
One line in the ADB report reads almost like a footnote to this mismatch:
Infrastructure needs should be understood as the infrastructure implications of regional growth and structural transformation.
Growth is shifting toward AI and digital services.
Economic structure is shifting toward data centers and computing clusters.
But the real infrastructure implications include things that are far less visible than headline investment numbers:
the sequencing of grid upgrades, queues for connection, and the rules governing cross-border electricity.
Those factors may ultimately determine how much of the US$7.1 trillion actually becomes usable computing capacity.
The money is on the balance sheet. The electricity is still on paper.
The gap between the two cannot be closed simply by adding another investment number.



