12 SEP 2026 — The Philippine government launched the final draft of its AI+ Infrastructure Masterplan 2026-2033 on 8 September, at a headline cost of 34.4 billion US dollars. It would take national data-centre capacity from a 50 megawatt baseline to 1.5 gigawatts. That is a thirtyfold increase in seven years, and 61 per cent of the money is expected from private investors who have not yet committed it.
A final draft, not yet policy
The status matters and is easy to lose. What was launched is a final draft, which is the stage at which a plan is complete enough to be argued about and not the stage at which anyone is bound by it.
The document is organised around six pillars: connectivity infrastructure, AI compute and data centres, sustainable energy and water sourcing, workforce development, policy and regulation, and demand creation and adoption. Two of those, energy and demand, are the ones the rest of this piece keeps arriving at, and it is to the plan's credit that both are named rather than assumed.
The numbers as published
Of the 34.4 billion dollars, 13.5 billion is public — 39 per cent — and 21 billion is private. Phase one targets roughly 400 megawatts by 2030, with the balance arriving in the three years after.
The plan names its geography. Clark-Bataan is the primary corridor, Batangas-Aurora the strategic gateway, with supporting hubs at Subic and in Calabarzon and regional nodes at Cebu, Iloilo, Davao and Cagayan de Oro. The projected returns are 500,000 AI-related jobs, a further 175,000 in infrastructure, and a 10 to 12 per cent uplift in GDP through productivity gains.
Which numbers are decisions and which are hopes
Three kinds of number appear in a plan like this, and they carry different weight.
The capacity targets and the corridors are decisions. A government can designate Clark-Bataan and can permit 400 megawatts, and those are things it controls.
The 13.5 billion in public funding is a commitment, subject to the usual qualification that a multi-year capital programme spans more than one administration and more than one budget.
The 21 billion in private investment and the 500,000 jobs are forecasts of other people's behaviour. At 61 per cent of the total, the plan's success depends on factors the government does not control.
The approval-to-arrival gap is the thing to watch
There is a useful comparison next door. Laos approved 34 investment projects worth 8.5 billion dollars in the first seven months of this year, a 68 per cent increase year on year. Actual capital inflows over the same period were 1.06 billion.
Roughly one dollar arrives for every eight approved. That is the ordinary shape of infrastructure announcements across the region, and not a Laotian peculiarity. We found the same pattern in a regional data-centre pipeline where announced capacity was not connected capacity.
This does not mean the plan will underdeliver. It means the numbers to watch are megawatts energised and dollars disbursed, not the megawatts planned and dollars pledged that make headlines.
Thirty times in seven years, in context
The thirtyfold multiple is the plan's real ambition, stated as a baseline rather than a boast.
Fifty megawatts is a small national data-centre footprint. For scale, single facilities announced elsewhere in the region are individually larger than the entire Philippine baseline, so the plan is not describing growth so much as a standing start. That is why the multiple looks extreme and why it is less extreme than it sounds: multiplying a small number by thirty still lands somewhere modest by the standards of Johor or Jakarta.
It also means the constraint is not demand. There is no serious question about whether 1.5 gigawatts of regional AI capacity will find customers. The question is whether this is where it gets built. Power price, permitting speed and connectivity decide that, and every competing market has a masterplan of its own.
Power is the binding constraint
Going from 50 megawatts to 1.5 gigawatts of data-centre load is a generation problem before it is a construction problem. Fifteen hundred megawatts of continuous demand is a significant share of any grid, and it must be firm, not intermittent, because a training run does not pause for cloud cover.
The plan acknowledges this by naming nuclear, exploring generation under Republic Act 12305, the PhilAtom Act. That is a serious answer to a serious constraint and it operates on a timescale of its own: a country with no operating civil reactor does not commission one inside the plan's window, so whatever powers phase one will not be nuclear.
The regional context makes this sharper, not softer. Singapore is importing 900 megawatts of solar from Johor; the most electricity-constrained market in ASEAN is solving its problem by buying capacity from a neighbour. Everyone in the region is bidding for the same firm power, and the Philippines is doing so from a starting position of higher electricity prices than most of its neighbours.
The jobs figure needs the most care
Five hundred thousand AI-related positions is the number most likely to be quoted and least likely to be checked, because "AI-related" has no standard definition.
Data centres are not employment-dense. A large facility runs with a few hundred staff. The separate 175,000 infrastructure jobs are mostly in construction — real work, but temporary.
The 500,000 must therefore come from firms using AI rather than from the facilities hosting it. That is a plausible forecast and a different proposition from the infrastructure the plan funds, because nothing about building 1.5 gigawatts in Clark-Bataan causes a Philippine company to adopt anything. The workers who would fill those roles are also the ones currently exportable, and a domestic AI sector competes for them against overseas employment at overseas wages.
What to watch
Phase one disbursement against phase one target. Four hundred megawatts by 2030 is specific enough to be checked, and the first two years of actual spending against the 13.5 billion public commitment will say more than any further announcement.
Watch for the first private investor to sign. With 61 per cent of the plan dependent on uncommitted capital, the first large private deal — its size, location, and power contract — is what turns this from a document into a project.
View author profile → · Editorial policy
