A trillion-dollar artificial intelligence company, according to the Bases Conversion and Development Authority (BCDA), spent two months studying New Clark City for a data center. It chose Johor, Malaysia instead. The reason given by the BCDA was electricity. There was not enough of it, and there would not be enough of it soon enough to matter.
That single rejection now explains a month of official contradiction. In July, the Department of Information and Communications Technology was telling reporters that modern data centers recycle their own water, that some desalinate and share the surplus with nearby towns, and that Secretary Henry Aguda was personally telling Amazon Web Services and Google to build in the Philippines instead of elsewhere.
By August 11, BCDA President Joshua Bingcang and Board of Investment Undersecretary Ceferino Rodolfo were saying, on the record, that no AI data center will be built at New Clark City at all, that the earlier framing was ‘just a misconception,’ and that the 1,620-hectare site is now zoned for semiconductor fabrication, microchip assembly, and the building of AI hardware components, not server farms. Bingcang added, almost in passing, that data centers do not generate many jobs anyway, a smaller and quieter claim than the hub-wide job figures Aguda had floated in July.
Manila is telling itself the same story as Aesop. The fox decided the grapes were sour only after it could not reach them. A government does not discover, unprompted, that hyperscale infrastructure was never really the plan. It discovers this after the investor it courted studies the grid for two months and leaves for a country that already has the power.
The revised framing is not simply the government saving face. Pax Silica’s stated purpose now includes securing supply chains for critical minerals and computing components among partner nations, which places the Philippines inside a contest larger than any single data center. China refines most of the world’s rare earths and critical minerals, and the countries that signed on to Pax Silica are, among other things, agreeing to build supply routes that do not run through Beijing.
A manufacturing hub making chip components is a genuine seat at that table. It is a smaller seat than the one the country wanted in July. But the difference between the two is the electricity Malaysia has and the Philippines does not.
The unnamed investor’s choice of country points at what Malaysia can now support. Racks Central, based in Singapore, is building Southeast Asia’s largest AI data center campus in Johor, a US$6.7 billion project needing 100 megawatts at first and 510 megawatts by 2028, drawing cooling water from a reservoir already built for the purpose and running it on solar, biomass, and hydropower already online. The capacity that made Malaysia the answer is not in question. The Philippines offered land, tax breaks, and a promise that power would arrive by 2028. Malaysia already had the power.
Multinational investment bank Morgan Stanley’s own numbers explain why that gap matters more now than it would have two years ago. A hyperscale facility running at full efficiency, in the bank’s arithmetic, rents US$25 billion of computing a year to sell only US$23 billion of output, and the bank expects the selling price of that output to keep falling as China’s DeepSeek open-weight models undercut the closed American labs on price. An exchange that lists oil and corn futures opens trading in GPU rental contracts on October 5.
When a product becomes a commodity with a public futures price, the sites that host it stop competing on hospitality and start competing on the one input that determines the cost of the commodity itself: electricity, delivered now, at scale, without qualification. A country building chip components for that commodity, rather than hosting the commodity’s production directly, is betting on a steadier and less capital-intensive corner of the same supply chain. This may be the wiser position, even if it happened by accident.
Manufacturing components for critical minerals and AI hardware is a real business and a defensible one, arrived at only after the larger prize walked out the door.
Nobody forced the company to leave. It did the arithmetic Malaysia’s grid could satisfy and the Philippines could not, and it left without waiting to see if Manila’s 2028 promise arrived on schedule. What the Philippines kept afterward, a manufacturing seat in someone else’s supply chain, is smaller than what it lost. It is also the only seat that was ever actually on the table. Whether the government can make even that much work is still unproven.
E-mail me at mangun@gmail.com. Follow me on Twitter @mangunonmarkets. PSE stock-market information and technical analysis provided by AAA Southeast Equities Inc.