Installed Is Not Dependable: A 123 MW Power-Planning Gap in Mindanao

A power figure is only useful when it answers the right question. For data-centre developers, investors and infrastructure buyers, the question is rarely “how much generation equipment exists?” It is “how much capacity can be relied on, under the conditions that matter to this site?”

That difference is visible in Mindanao. A direct PowerlandMap extraction on 25 October 2026 covers 24 operating diesel and oil-based plants documented by the Philippines Department of Energy (DOE). Together they represent 727.4 MW of installed capacity and 604.4 MW of dependable capacity. The gap is 123.0 MW, or 16.9% of nameplate capacity.

This is not a claim about the whole Mindanao grid, the availability of any specific connection, or the capacity that could serve a data centre. It is a bounded comparison of 24 plant records imported from the DOE’s May 2026 plant list. But it demonstrates a decision rule that I think should be standard in digital-infrastructure screening: installed megawatts are a starting point, not a deliverability conclusion.

The PowerlandMap evidence

PowerlandMap reconciled plant identity, operator, locality, technology, operating status, installed MW and dependable MW against the DOE’s List of Existing Power Plants as of 31 May 2026. No coordinates were inferred where the source did not provide a verified point. Black-start engines and battery systems were not converted into extra generation capacity. Where the unit count was not legible, it remained unknown.

The resulting comparison is:

The DOE has long described installed capacity as the maximum electricity a plant can produce based on manufacturer-rated equipment. Its monthly summaries separately publish installed and dependable capacity. That separation is not cosmetic. It prevents a model from treating equipment rating, operational availability and deliverable grid service as if they were the same thing.

The distribution also matters. The 123 MW gap is not spread evenly. Iligan Diesel Power Plant – Plant 2 accounts for 72.5 MW of the difference on its own: 114.4 MW installed versus 41.9 MW dependable. Western Mindanao Power Corporation’s Sangali plant has a 10.3 MW gap. Southern Philippines Power Corporation’s plant has a 6.7 MW gap. Other facilities are much closer to nameplate, and some show equal installed and dependable values.

A country or regional average therefore cannot replace facility-level diligence. Two sites may sit in the same power market but face very different upstream generation, network and contractual conditions.

The decision for a data-centre shortlist

The practical decision is whether a location should advance from market screening into connection, land and power diligence.

My threshold would be simple: do not advance a site on installed-capacity evidence alone. Before allocating material engineering or acquisition time, require three separate answers:

1. What generation is registered in the relevant system? 2. What capacity is dependable under the latest official reporting? 3. What capacity is contractually and physically deliverable to the proposed connection point, on the required date?

The first two questions can be supported by a plant register. The third cannot. It requires a utility or system-operator process, substation and transmission evidence, queue position, protection studies, outage assumptions and commercial terms.

This is where data-centre market intelligence often fails. A regional capacity total is useful for orientation, but it does not prove that a campus can secure firm service. Similarly, a power-plant location near a development parcel does not prove electrical adjacency. The transmission topology, voltage level, contingency standard and connection schedule remain distinct evidence layers.

The National Grid Corporation of the Philippines describes its Transmission Development Plan as the blueprint for expanding and improving the network. Its project pages explicitly connect transmission investment with reliability and security. That is the correct analytical layer for the next stage of screening, but a plan is still not a site-specific connection offer.

A transparent sensitivity test

Consider a hypothetical 20 MW IT deployment. This is not a forecast for Mindanao and not an estimate for any project. It is a reproducible way to show how planning assumptions interact.

Assume:

• 20 MW of IT load
• PUE of 1.30
• 15% planning contingency above site load
• the 83.1% dependable/nameplate ratio observed in this bounded plant sample, used only as a sensitivity proxy

The calculation is:

• 20 MW IT × 1.30 = 26.0 MW site load
• 26.0 MW × 1.15 = 29.9 MW planning requirement
• 29.9 MW ÷ 0.831 = 36.0 MW nameplate equivalent

The sensitivity does not mean that a data centre can reserve part of this plant portfolio. It shows the modelling error created when nameplate capacity is treated as dependable. For the 20 MW IT case, using installed capacity as if it were dependable would understate the proxy requirement by about 6.1 MW before considering grid losses, maintenance, fuel constraints, reserves, transmission limits or commercial availability.

The counter-argument is that a diversified grid does not depend on one plant or one fuel type. That is correct. System operators manage portfolios, reserves, outages and network constraints across a much larger fleet. A site should not be rejected merely because individual diesel units have a nameplate gap. The alternative wins when the developer can show firm utility service, credible network reinforcement and a delivery schedule that absorbs those system risks.

That is precisely why the plant-level metric should be used as a screening signal, not as a bankability verdict.

What the dataset does not prove

The current PowerlandMap extraction is intentionally narrow.

It does not prove:

• available capacity at a named substation
• firm or interruptible service terms
• fuel availability or dispatch economics
• hourly or seasonal production
• N-1 or N-2 network compliance
• the timing of transmission upgrades
• the carbon intensity of a proposed supply contract
• data-centre IT capacity
• customer demand or a signed project

The sample also uses one aggregated PowerlandMap unit record per plant because the DOE source does not provide a safely reconciled per-unit capacity split for every facility. The article therefore counts plants and aggregate records, not physical generating sets.

The next useful enrichment is to connect each shortlisted demand location to verified substations, transmission projects and utility processes, while preserving the distinction between grid-connected and embedded generation. That work must not infer a route from geographic proximity alone.

A better power-readiness workflow

For a developer or investor, I would structure the evidence pack in five layers.

Layer 1: registered generation. Identify operating plants, fuel, operator, installed MW and dependable MW from the latest official register.

Layer 2: network topology. Verify substations, voltage levels, transmission corridors and approved reinforcement projects from system-operator sources.

Layer 3: connection status. Obtain the formal application position, study stage, requested load, delivery date and unresolved technical conditions.

Layer 4: commercial firmness. Separate an indicative conversation from a signed connection agreement, and separate firm service from interruptible or conditional supply.

Layer 5: operating resilience. Test outage exposure, backup generation, fuel logistics, UPS duration, maintenance assumptions and recovery procedures against the workload.

PowerlandMap can make the first two layers faster to investigate and easier to compare. It cannot replace the utility, engineering and contractual evidence required for layers three to five.

The conclusion

The 24-plant Mindanao sample contains 727.4 MW of installed capacity but 604.4 MW of dependable capacity. The 123 MW difference is large enough to change a screening model, yet concentrated enough to show why a single regional haircut is also inadequate.

The decision is not to discount every megawatt by 16.9%. It is to stop using nameplate capacity as a proxy for site-ready power.

For every serious data-centre shortlist, keep installed, dependable and deliverable capacity in separate fields. Reconcile the source date. Preserve unknowns. Then require site-specific network and commercial proof before presenting power as secured.

Sources and method

• Philippines DOE — Mindanao List of Existing Power Plants, 31 May 2026
• Philippines DOE — regional capacity summary, 31 May 2026
• Philippines DOE — Electric Power Industry Management Bureau
• NGCP — Transmission Development Plan 2024–2050
• NGCP — Operations and transmission projects

PowerlandMap method: production SQL read on 25 October 2026; 24 published non-demo plant records with source_system = ph_doe; installed MW summed from published aggregate unit records; dependable MW transcribed from the same DOE plant rows and reconciled separately; duplicates checked by external reference. Coverage is limited to the currently reconciled sample and should not be treated as the complete Mindanao power system.