Silicon dreams in a chapel

By Francis Allan L. Angelo

ILOILO CITY — The question that filled the chapel in Barangay Abilay Sur on Aug. 14 was what a 50-megawatt artificial intelligence data center would do to Oton, particularly to its water, power, and Its farmland.

But the bigger elephant that nobody bothered to mind was: could a facility of that description operate in Oton at all?

Impact analysis assumes the thing gets built and then runs for 20 years. Based on the evidence presented across 18 slides and roughly 90 minutes of raucous consultation, that assumption is the least examined part of the proposal, and it determines whether any of the rest matters.

SITE INPUTS

A hyperscale data center is a business with two hard requirements: firm electricity at a competitive price, and low-latency fiber to the customers it serves. Everything else – land cost, labor, tax holidays, goodwill – is secondary. Site selection in this industry is unsentimental, and Oton is weak on both counts.

Let’s start with power. Philippine industrial electricity has long been among the most expensive in Southeast Asia, and the incentive package assembled to close that gap – the CREATE MORE deductions on power costs, reduced corporate income tax for export-oriented locators, and the green lanes created under Executive Order 18 – was modeled largely around the Meralco franchise. Oton is not included because Iloilo I Electric Cooperative Inc. serves it on an island sub-grid where congestion is already pushing costs up. Aurora Energy Research told a forum in Iloilo City in June that Panay and Negros are among the Visayas areas where the system is signaling a need for more generation and stronger grid capacity, and that the Visayas grid is not robust enough to optimize around a single plant going offline.

A flat 50-megawatt load placed at the end of a congested transmission line does not buy the cheap version of Philippine electricity. It buys the expensive version, permanently, and it competes for tenants against sites in Cavite, Batangas, Pampanga, and Tarlac that do not carry that penalty.

Now the fiber. The international submarine cables that make a Philippine data center useful to a global hyperscaler come ashore at Batangas, Nasugbu, Cavite, La Union, Daet, Baler, and Digos. Iloilo is served by domestic backhaul — the Philippine Domestic Submarine Cable Network, Converge’s domestic system, and the PLDT and Globe backbones. That is entirely adequate for business process outsourcing, enterprise hosting, and government workloads. It is not the same thing as sitting on an international landing point, and the difference shows up as latency and as a recurring backhaul lease that a Luzon competitor does not pay.

This is not an abstract objection. It’s why Philippine pipeline clusters are where they are, and why the country’s largest project under construction — the Globe, ST Telemedia, and Ayala campus in Fairview scaling toward 124 megawatts — sits inside the Meralco franchise rather than in a province.

UPTIME IS THE CONTRACT

Hyperscale tenants do not rent floor space. They buy an availability guarantee, and the operator bears the financial penalties for missing it.

Panay lost power island-wide for four days in January 2024. Data presented to the Energy Regulatory Commission showed 3,095 reported outages in the Visayas from January to June 2026, the highest among the three major Philippine grids for that period. That figure describes the grid, not any data center, and no causal link between the two has been established. But it describes the conditions under which an operator in Oton would have to sell an uptime commitment.

Diesel generators and battery systems are designed to bridge minutes and hours. They are not designed to carry a 50-megawatt continuous load through a multi-day island system collapse, and the fuel logistics of trying would be its own line item.

The Department of Energy is currently drafting a Panay Energy Master Plan, citing the island’s position as the end of the line in the Visayas transmission system and a supply risk flagged for 2029. The proposal reaches full capacity inside that same window.

THE SUBSTATION QUESTION

One version of this project would address most of the above, but it was not the version presented. If a developer arrives with its own generation, such as dedicated solar, a gas plant, contracted baseload under a bilateral supply agreement, then the grid constraint becomes a much smaller problem.

Asked about the project’s electricity requirement, Robertsen said he would not draw on the community’s supply. “I will not connect to use your power. I will build my own power substation on the land,” he said, describing a developer-funded facility with a 69-kilovolt to 115-kilovolt interconnection and dual-feed redundancy.

But a substation does not generate electricity. It steps voltage up or down so power can move between the transmission network and a customer. What was described is a dedicated connection that routes around ILECO I’s distribution network, not around the Panay grid. Every megawatt consumed in Abilay Sur would still come from the same generators that supply the rest of the island, drawn through the same constrained corridor.

The letter of intent sent to ILECO I concerns delivery. Not one slide presented at the consultation addressed where the 50 megawatts would be generated, or under what supply contract, and that is the single largest gap in the feasibility case.

WHAT IS THE SELL?

Feasibility questions of this kind have a standard answer in the global industry, and it is worth understanding before assuming bad faith.

The companies that operate hyperscale computing build comparatively few of their own buildings. Between them and the ground sits a layer of developers whose product is not a data center but a site: land with zoning cleared, an interconnection secured, water arranged, and permits in hand. The industry term is powered land. It is sold or leased to whoever eventually installs the servers.

That layer has a documented failure mode. In the United States, practitioners estimate that developers file five to 10 times more large-load interconnection requests than there are facilities actually built, a pattern utilities now call phantom data centers. Exelon has said only about 22 percent of its 65-gigawatt pipeline through 2040 is likely to materialize. CenterPoint Energy in Houston watched its data center interconnection requests climb from 1 gigawatt to 25 gigawatts inside 12 months.

The mechanism is more optionality than fraud. A queue position, a letter of intent, and an option on land cost very little and can be sold across several jurisdictions at once. The developer builds where the power price and permitting are best, abandons the rest, and the abandoned proposals still distort the planning of every utility that took them seriously.

The markers are recognizable. A letter of intent rather than a supply contract; no named anchor tenant; a phased build that starts at 10 megawatts and promises 50; a single principal rather than an institutional balance sheet; thin corporate documentation; a build schedule – 15 months to first capacity – that assumes interconnection studies; an environmental compliance certificate: and cooperative and regulatory approvals move faster than they have moved for anyone else on Panay.

The true scope of the Oton proposal remains unclear. The chapel hearing exposed only the questions that must be answered, while providing zero answers.

WATER FIGURE

The number that caught the most attention after the consultation was 2,500 cubic meters of water per day at full capacity. Robertsen was right to resist that framing. The figure he gave describes the volume held inside a closed cooling loop, not water withdrawn and destroyed each day. Closed-loop and liquid direct-to-chip cooling are real technologies, and they are meaningfully less thirsty than the evaporative towers that produced the alarming headlines out of Arizona, USA and Chile.

The number that matters is makeup water: the volume that must be replaced because it evaporates or is bled off to control mineral buildup. Robertsen conceded that evaporation would occur, but he did not say how much. That single figure, expressed in cubic meters per day at 50 megawatts, is what Metro Pacific Iloilo Water and the Oton council would need to assess whether the project competes with households during a dry season. It was not in the 18 slides Robertsen presented, and it was not established in the question-and-answer.

For scale, Bloomberg has reported that a typical 100-megawatt data center can consume roughly 2 million liters of water a day. Whether Oton’s proposal would sit anywhere near that range is unknown, because the one disclosure that would answer it was never made.

THE MISSING GATE

After 15 months, the proposal faced its first real test: a barangay chapel with a projector. We are not mocking the town of Oton – it’s just how the Philippine permitting system works. Barangay consultations, Sangguniang Bayan endorsements, ECCs, and zoning clearances are real hurdles, but they scrutinize only the project itself. They completely ignore the proponent. At no point does the bureaucracy demand an audited balance sheet, a signed offtake agreement, proof of site control, or even a refundable deposit for the grid study the developer expects a cooperative to run.

Jurisdictions that absorbed a wave of speculative proposals built those gates afterward and funded them. ComEd in Illinois now charges USD 1 million to anyone requesting 50 megawatts or more. Ohio requires new data centers to pay for at least 85 percent of their projected consumption whether they use it or not. Virginia locks large-load customers into 14-year contracts. Texas requires large loads to fund infrastructure upgrades and to disclose duplicate applications filed elsewhere. The effect was immediate and measurable: developers added roughly half as much pipeline capacity in one quarter as in the preceding one.

The Philippine policy stack currently points the other way, and deliberately so. Executive Order 18 created green lanes to accelerate strategic investments. The Data Center Association of the Philippines has set a target of 1 gigawatt of capacity by 2029, and the Department of Information and Communications Technology floated an 18-gigawatt expansion over the next decade. The country has assembled a full set of instruments for saying yes quickly and almost none for establishing who is asking.

BEYOND ONE WALKOUT

Iloilo has the profile the siting model hunts for: cheap coastal land immediately beside a fast-growing city, a business process outsourcing base that reads as digital readiness, an ordinary distribution cooperative that has never had to price a 50-megawatt flat load, and a permitting path where the first hard question arrives at the barangay level. Whatever happens to this particular proposal, the profile does not change. Others will come, and the more credible ones will arrive with lawyers and a communications plan instead of a temper.

The Sangguniang Bayan is expected to take up a committee report on the proposal this month. The most useful thing it could produce is not a resolution about one man’s conduct but a standing checklist: what a large-load proponent must file before a public consultation is scheduled.

The information that would have settled most of the questions raised in Abilay Sur is neither exotic nor proprietary. It is a signed offtake agreement, an audited balance sheet, a grid impact study commissioned at the developer’s cost, and a makeup-water figure.

Unsurprisingly, the bureaucracy doesn’t force a public hearing to uncover those details, and the developer certainly did not flash them on the chapel projector.