Article 9: Who Pays for the Power Behind AI?
The Philippines wants 1.5 GW of AI-ready data centers. Before we build the grid around that ambition, we need rules for who pays for what.
The Philippines has set an ambitious target.
Under the Philippine AI+ Infrastructure Masterplan 2033, the government is targeting 1.5 GW of AI-ready data-center capacity and about US$30 billion in investment.
That sounds like a technology story.
It is also an infrastructure-financing story.
A 100 MW or 300 MW hyperscale data center cannot necessarily connect wherever land happens to be available. Depending on the location and existing network capacity, accommodating such a load may require transmission reinforcement, new substations, dedicated connection facilities, additional generation contracts and significant on-site electrical infrastructure.
Someone has to pay for those investments.
And that leads to a question the Philippines should answer before hyperscale development accelerates:
Who caused the cost, who benefits from the investment, and who should therefore pay?
Not all electricity infrastructure is the same
This distinction matters.
The electricity supporting a hyperscale data center involves several layers, and their costs should not automatically be treated alike.
Generation may be procured commercially through power-supply arrangements or the electricity market.
Shared transmission infrastructure may serve the data center while simultaneously strengthening the wider grid.
Dedicated connection facilities and substations may exist primarily because the hyperscale customer requires them.
On-site electrical systems, backup power and storage principally serve the facility itself.
The policy challenge is therefore more sophisticated than saying either “make Big Tech pay” or “let the grid absorb the investment.”
The correct allocation depends on why the infrastructure became necessary and who receives its benefits.
Cost causation first
Imagine a developer proposing a 300 MW AI campus.
The existing transmission network cannot accommodate it. A new connection facility and major network reinforcement are required.
The first question should be:
Would this infrastructure have been required if the 300 MW data center had never arrived?
If the answer is no, there is a strong case that the new customer should bear much of the incremental cost it caused.
But there is a second question.
Suppose the transmission upgrade also improves regional reliability, unlocks renewable generation and provides additional capacity for future industries and communities.
Now the infrastructure has beneficiaries beyond the data center.
That suggests a more defensible principle:
Cost causation first. Beneficiary analysis second. Cost allocation follows.
Figure 1. A 300 MW AI Data Center Arrives: Who Causes the Cost, Who Benefits, Who Pays?
The proposed framework distinguishes generation, shared transmission, dedicated connection facilities and on-site infrastructure before determining how costs should be allocated.
Could data centers actually benefit consumers?
This is an important counterargument.
Large data centers are enormous electricity customers. Their arrival can increase electricity sales, support investment in new generation and potentially spread some fixed system costs across a larger customer base.
They may also help justify transmission projects that improve reliability or unlock generation resources that benefit other users.
So the problem is not that data centers require infrastructure.
The problem arises when costs allocated to existing consumers are greater than the benefits those consumers receive.
That distinction matters because the Philippines should neither subsidize private hyperscale infrastructure indiscriminately nor create rules so punitive that they discourage investment that could genuinely strengthen the electricity system.
The objective should be fair allocation.
Other countries are already confronting this
Virginia, home to the world’s largest concentration of data centers, is wrestling with precisely this issue.
Regulators have sought mechanisms intended to reduce the risk that transmission costs associated with very large data-center loads are shifted to other electricity customers. Microsoft is challenging one such ruling involving upfront transmission infrastructure costs.
Texas presents a different warning.
The problem there is not simply how much electricity data centers may eventually consume. It is how much prospective demand developers are asking grid planners to reserve.
Texas alone has accumulated roughly 474 GW of large-load connection requests, while reported requests across several major U.S. power systems have risen beyond 700 GW. Much of that prospective demand is associated with data centers.
Not all those projects will be built.
The problem of “ghost demand”
A developer considering several locations may submit connection requests at multiple sites even though only one project eventually proceeds.
Other proposed projects may lack financing, committed customers or sufficient development maturity.
Yet grid planners must somehow decide which requests represent genuine future electricity demand.
Get Franz F. Ziebert, PMP®’s stories in your inbox
Join Medium for free to get updates from this writer.
That creates ghost demand: prospective loads appearing in connection queues that may never become operating facilities.
The danger is obvious.
If grid planners treat every speculative 300 MW request as certain demand, billions could eventually be committed to infrastructure that turns out to be unnecessary.
The Philippines has an opportunity to learn from this before the problem arrives here at scale.
Large-load applicants should therefore demonstrate meaningful project maturity before hundreds of megawatts are incorporated into long-term grid planning.
That might involve financial commitments, land control, development milestones and staged capacity reservations.
A request for 300 MW should be treated as an infrastructure commitment, not simply an application for electrical service.
The Philippine regulatory question
The issue is already becoming relevant here.
The Philippine Energy Efficiency Alliance has warned that rapid data-center expansion could strain grid capacity and potentially increase electricity costs without appropriate planning and policy.
The Philippine transmission framework is also evolving.
In June 2026, the Energy Regulatory Commission issued Resolution №18, Series of 2026, establishing rules under which qualified entities other than the transmission network provider may finance and construct certain Associated Transmission Projects and Priority Projects.
The resolution does not answer the data-center cost-allocation question.
Its significance is different.
It demonstrates that the Philippines is already reconsidering how certain transmission infrastructure can be financed and delivered.
Very large AI loads introduce a related issue that deserves explicit regulatory attention.
Establish the rules before the megawatts arrive
The Philippines should welcome hyperscale investment.
Data centers can attract capital, expand digital infrastructure, create demand for renewable energy and support a broader technology ecosystem.
But investment incentives should not quietly become infrastructure subsidies whose costs eventually appear on somebody else’s electricity bill.
Large-load policy should therefore answer two fundamental questions.
First:
If a project causes infrastructure that would otherwise not have been necessary, how much of that incremental cost should the project bear?
Second:
If that infrastructure also produces measurable benefits for the wider grid, how should those benefits be reflected in the allocation of costs?
Those questions should be answered before projects are approved and transmission investments committed.
Investors would gain greater certainty.
Electricity consumers would gain protection against inappropriate cost shifting.
Grid planners would gain better information about which projects are genuinely likely to proceed.
The infrastructure question behind AI
Our earlier articles have argued that the AI economy is ultimately a physical economy.
It requires land, water, electricity and digital connectivity.
But underneath those requirements lies another layer: Capital.
Someone finances the generation.
Someone builds the transmission infrastructure.
Someone pays for the substations.
And someone ultimately carries the risk if the projected demand never materializes.
As the Philippines pursues 1.5 GW of AI-ready data-center capacity, we should not wait until major transmission projects are already underway before deciding how those costs should be allocated.
The objective is not to make data centers pay for everything.
Nor should ordinary electricity consumers automatically pay simply because an infrastructure project becomes part of the wider grid.
The better test is straightforward:
Who caused the cost? Who receives the benefit? Who should therefore bear it?
The Philippines has an opportunity to establish those rules before hyperscale demand arrives.
We should design the economics before we build the grid around the demand.
References
Department of Information and Communications Technology. (2026). Philippine AI+ Infrastructure Masterplan 2033, as reported by GMA News Online, June 9, 2026. The plan targets 1.5 GW of AI-ready data-center capacity and approximately US$30 billion in investment.
Energy Regulatory Commission. (2026). ERC Greenlights New Transmission Rules to Support Energy Security. Resolution №18, Series of 2026.
Philippine Energy Efficiency Alliance. (2026). PE2 warns data center expansion could strain grid, raise power costs.
Reuters. (2026, September 1). Reporting on large-load connection queues, data-center demand and the emerging problem of speculative or “ghost” demand in U.S. electricity markets.
Financial Times. (2026, September 4). Reporting on Microsoft’s challenge to Virginia’s treatment of transmission infrastructure costs associated with large data-center loads.
Franz F. Ziebert, PMP®
Managing Director, Design Coordinates, Inc. (DCI)
Civil Engineer | Executive MBA | Project Management Professional
Doctor of Engineering Management Candidate
Engineering the Philippine AI Economy is a continuing series examining the infrastructure, energy, resources, sustainability and engineering decisions behind the Philippines’ emerging AI economy.
The views expressed are personal and intended to contribute to the continuing discussion on how the Philippines can build a competitive, resilient and sustainable AI infrastructure ecosystem.
