President Marcos signed Memorandum Order No. 47 on May 21, approving the 2026 Strategic Investment Priority Plan and placing artificial intelligence, cybersecurity, and data center facilities in Tier III, the plan’s highest incentive category, alongside quantum computing and advanced nuclear and hydrogen energy. The order was published in the Official Gazette on June 2 and took effect 15 days later, formally making AI infrastructure one of the handful of technologies the Philippine government now treats as critical to the country’s long-term economic competitiveness.
Tier III status matters in concrete terms: it unlocks the deepest available tax and non-tax incentives under the CREATE MORE Act, the same expanded incentive package that’s already been used to court manufacturing and export-oriented investment, rather than the more generic incentives lower-tier industries receive. The government’s own framing places AI and data center facilities in the same strategic bracket as frontier technologies most countries in the region are still years from formally prioritizing at all.
The building is already underway, separate from the widely reported $10 billion US-backed AI hub taking shape near Clark Freeport Zone. PLDT’s data center arm, VITRO, is designing a 100-megawatt facility estimated to cost roughly 40 billion pesos. ST Telemedia Global Data Centres is adding 12 megawatts of capacity in 2026 alone, en route to making its Fairview campus the country’s largest data center at 124 megawatts, representing roughly $1 billion of investment on its own. Neither project depends on the Clark hub materializing, both are domestic infrastructure bets already in motion.
The Clark hub itself has taken clearer shape since first being reported. It’s envisioned as the first physical facility under Pax Silica, a Washington-led initiative to build secure AI and semiconductor supply chains among allied nations, spanning 4,000 acres and intended to host AI computing infrastructure, semiconductor packaging, critical minerals processing for nickel, cobalt, and copper, and advanced manufacturing and logistics. US companies are expected to commit more than $1 billion in the hub’s early phases alone, including digital upskilling programs for millions of Filipino workers, and Manila has an option to offer Washington a lease running as long as 99 years. That scale of commitment has also produced friction: Manila has reportedly rejected US requests for diplomatic immunity over the site, insisting the facility remain under Philippine jurisdiction, a sovereignty line the government has held even while racing to finalize an initial agreement with the US State Department this year.
The scale of the shift is genuinely fast by regional standards. The Philippines’ total data center capacity is projected to roughly triple, from about 150 megawatts currently to nearly 500 megawatts by 2028, before accounting for whatever capacity the Clark hub eventually adds on top. Established players, ePLDT, Converge ICT, Globe, Digital Edge, and Equinix, are all expanding alongside newer entrants like A-FLOW and Digital Halo, while Huawei Cloud and Alibaba Cloud are both expanding local availability zones. The overall colocation market is projected to climb from roughly $735 million in 2025 toward $2.48 billion by 2031.
The bottleneck underneath all of this is power, not capital or government incentive, and the numbers involved are large enough to matter well beyond the tech sector. Industry estimates put ICT-sector baseload power requirements at 300 to 500 megawatts by 2026 alone, on top of ordinary economic growth in demand, a figure the country’s existing generation and transmission build-out was never sized for. Wholesale electricity prices across Luzon, Visayas, and Mindanao surged 58 percent in March 2026, and grid reserve margins remain thin enough that the national grid operator has flagged continued vulnerability to unplanned plant outages and demand spikes through mid-2026. Meralco is responding at the individual-project level rather than waiting for system-wide fixes: it commissioned a dedicated 115-kilovolt switching station to serve ePLDT’s VITRO facility in Santa Rosa, energized a dual-feed station inside ST Telemedia’s Polaris campus in Quezon City to guarantee stable power to that roughly 180-megawatt site, and has told researchers it expects to deliver 1,000 megawatts combined to ten data centers in the coming years. None of that changes the underlying reserve-margin math for the grid as a whole, it just prioritizes power delivery to the specific facilities large enough to negotiate their own dedicated infrastructure.
Every one of these projects, in other words, is competing for grid capacity in a country where electricity pricing and supply reliability are already contentious political issues entirely separate from AI. Tripling data center demand for power lands on a grid that was never sized with AI-scale compute density in mind, and none of the incentive tiers in the SIPP directly address that constraint, they lower the cost of building, not the cost or availability of the electricity needed to actually run what gets built.
For Philippine AI startups, the practical question the SIPP incentives raise isn’t about data center ownership, almost none of these projects are being built for cheap compute access by local startups, they’re being built for hyperscaler and large enterprise colocation tenants capable of filling a 100-megawatt facility on their own, or dedicated national-security infrastructure under an agreement Manila is still negotiating jurisdiction terms for. The more relevant question is whether this new local supply eventually pushes GPU-hour pricing down for smaller Philippine tenants at all, or whether, as with the Clark hub before it, the capacity gets absorbed entirely by foreign enterprise and government clients before it ever meaningfully reaches a Series A Philippine AI startup’s monthly compute bill.
The SIPP’s tax incentives make the Philippines a cheaper place to build AI infrastructure than it was a year ago. They say nothing yet, on their own, about who actually gets to rent time on what gets built, or whether the power grid underneath it can keep pace with a tripling of demand on top of an already-strained baseload, and both of those distinctions are likely to matter more to the domestic startup ecosystem than the headline capacity figures being reported right now.
Share this article