Taiwan: PGIQ Rating, Tier 4 Constrained

The same market, read four ways

BuildSellPlanMonitor

Taiwan is a limited market to sell power into

Grade C (Limited)

Offtake score 44/100

Demand 55, build feasibility 22, combined into the score. For a seller: where buyers already pay, how tight supply is, and how feasible new build is here.

Price to beat

A 100 MW load on our standard case pays about $133/MWh delivered here. A competitive supply offer works against that level.

Connection

Connection difficulty Severe, typical wait 3-5 years.

See it on the Offtake Grade →

Taiwan: effectively full at system level, rising demand pressure

Effectively full 18/100

System-level room, not local capacity

Local capacity: not assessed. We have found no operator publication of area or substation capacity for this market, so whether the bus serving a candidate site can take the load is an open question. Absence of a finding here is not evidence that capacity exists.

Two reads for a grid planner: how much room the system as a whole looks to have (Effectively full), and how hard demand is already pushing (Rising). The first is a market-level estimate from connection ease, forward adequacy and reliability margin. It is not local deliverability. Capacity at a specific substation is not assessed for this market, and a market can look roomy while the bus you want has nothing available. Confirm local capacity with the utility before relying on it.

Demand growth

Moderate load growth in our read, the pull on your system.

Load Pipeline Reality → Where demand is hungry → Connection friction →

Grid Risk Annex BETA

Watch

Needs structuring or mitigants · grid strength 48/100

A cited, dated grid-risk read on Taiwan for a capital provider, the banks and private-credit lenders, infrastructure funds and insurers holding or underwriting a power or data-centre exposure here. Drop it into a credit memo and watch each factor over the life of the exposure.

Read it against your asset. For a load asset (Buy), weight connection access and energization timing. For a generation or offtake asset (Sell), weight offtake demand, curtailment (when a plant is told to stop feeding the grid) and cost realisation. For a grid or transmission asset (Move), weight load-pipeline reality and buildout risk.

Rating and track record

PGIQ Tier 4 Constrained, outlook Stable, Med confidence. Scored on a public, dated, falsifiable record you can cite in a credit memo.

Delivery reliability

Minor flags found in the bulk and local outage record we reviewed.

Future firm supply

Reviewed; firm supply ample.

Cost certainty

Recurring cost about $133/MWh (band $128 to $140) for the standard 100 MW case, basis administered. Excludes connection capital.

Regulatory / stroke-of-pen

Not assessed.

Execution risk

Connection friction Severe, the risk the project or load actually energizes on time.

Ongoing monitor

Track this market's rating actions and the Global Power Index as a covenant-style monitor over the life of the exposure.

Monitoring plan

Covenant-style triggers to watch over the life of the exposure: a change in the PGIQ rating or outlook; a dated regulatory move (stroke-of-pen risk); a capacity-auction or cost-band shift; a forward firm-supply adequacy warning; and a rise in queue drop-out or connection friction. Track them on this market page and the Global Power Index, or as Analyst Desk alerts.

Sources: Regional grid data

This is a directional grid-risk screen, not a credit rating, and not investment, engineering or procurement advice.

The formatted, cited Grid Risk Annex export is an Analyst Desk feature.See the Grid Exposure Monitor →

The five pillars behind the tier

The pillar signature: Access, Availability, Cost, Momentum, Carbon, taller is stronger. The same shape repeats across every market so peers compare at a glance.

Access30%45

Constrained access; TSMC competes for power.

Availability25%40

Tight availability on an island grid.

Cost25%25

Very high cost.

Momentum15%45

Moderate momentum.

Carbon5%43

Carbon moderate-high (~400 g).

Default weights, which lean toward cost and access rather than carbon: Access 30 · Availability 25 · Cost 25 · Momentum 15 · Carbon 5. Momentum reflects the data-center build already underway, not just stated intentions.

Reliability overlay

Reliability adjustment: the five-pillar score of 39 is reduced to 35, a deduction of 4 points, reflecting the delivery-reliability and firm-supply risks flagged below. Reliability is reported as five separate layers, never as one combined label, because they answer different questions and a strong answer on one does not cover a gap in another: what has actually failed on the bulk system, what a forward assessment expects of firm supply, how the local network performs, what your contract actually promises, and what your site is designed to ride through. Any one of them can cap the tier.

Historical bulk-system performance: Minor flags found

Modelled read: a tight island grid with priority industrial load and recent large outages (2021-2022) and thin reserve.

What this means: A large new load faces occasional exposure here. Monitor conditions and keep a contingency plan.

modelled

Historical utility-territory delivery performance: No reliability flags found

Forward resource adequacy (assessed): No adequacy flags

Contractual service terms: not assessed

We have not compiled what firm service actually promises here: whether it is firm, non-firm or interruptible, the conditions under which you can be curtailed, the redundancy required of you, and the transfer time behind any N-1 commitment. Treat this as an open diligence item, not as an indication that service is firm.

Ask the utility for the tariff or service agreement terms before relying on firm supply.

Designed site resilience: not applicable at market level

Utility feeds, UPS, storage, on-site generation, islanding, black start and the residual energy you still expect to lose are properties of a specific site design, not of a market. This layer only exists once there is a site, and we never assert it for you.

Confidence: Low. Each input is labelled as measured, modelled, or drawn from a specific event, and is framed as a risk to a prospective large new load, not as a verdict on any utility or grid operator. Forward adequacy is an assessment of what is expected, not a measurement of what has happened; it is a forecast and is labelled as one.

Connection friction

How hard it is to connect a large new load here, and how long it takes. Part of the PowerGridIQ Connection Friction Feed. Labelled proxy; confidence Med.

Friction level

Severe

Typical connection wait

3-5 years

Source: CAPRI: the energy paradox of Taiwan's AI ambition

This market's friction level, wait band and full story are free here. The precise typical wait in months, and every market ranked side by side, are the Analyst Desk.

Taipower has paused new 5 MW-plus connections north of Taoyuan; grid interconnection runs 3 to 5 years.

Grid interconnection in Taiwan typically runs three to five years, and Taipower has restricted new 5 MW-plus data-centre connections north of Taoyuan unless new capacity is built, because demand clusters in the north while generation sits in the south over a corridor not designed for it. Transformer shortages add six to twelve months; four new 161 kV substations due 2025 to 2027 aim to ease it.

What builds the 54-month wait

Application & Taipower review12 mo

Grid study & connection offer12 mo

Substation & network reinforcement (161 kV)18 mo

Construction & energization12 mo

Typical time to energization54 months

How the typical connection wait breaks down by stage. Stages overlap in practice and vary by project; the split is our estimate anchored to the operator process, and the total is the representative time to energization. Anchored to the operator process (CAPRI: the energy paradox of Taiwan's AI ambition).

Data-center share of demand pressure: High. Sources: CAPRI: the energy paradox of Taiwan's AI ambition · Taipower tiered data-centre rates (DigiTimes). The full cross-market connection dataset is available through the Connection Friction Feed.

Large-user tariff

This market has no competitive wholesale power market, so the honest cost is the administered tariff a large load actually pays, all charges included, not a modelled spot price. A tariff market needs no live feed: it changes only when the regulator changes it.

Industrial tariff

NT$4.27 per kWh (about $0.133 per kWh)

administered, Taipower industrial rate, 2025/26

Trend

Frozen for six months through March 2026 to curb inflation, with Taipower absorbing rising fuel costs that could lift it later.

Taiwan's power comes from state utility Taipower at an administered industrial rate of NT$4.27 per kWh (about 13 US cents), frozen through March 2026. There is no wholesale market; the tariff is what a large load pays, and the state is absorbing fuel-cost pressure for now.

Basis: measured. Source: Taipower industrial rate (Taipei Times).

Realized cost band

What a large electricity user pays for power here on one declared case, 100 MW contracted demand, 95% load factor, HV service, per megawatt-hour. Not the wholesale price and not the sticker tariff: energy plus the grid, delivery and other charges that land on the bill. It is a range, because real contracts vary, and it is labelled by how it was built.

Recurring cost, delivered

$128 to $140 per MWh

Midpoint about $133 per MWh

How this band was built

administered

From 1 public fragment

What sets this band

There is no wholesale spot market here, so the band is not an energy-plus-network stack. It is set by the regulated tariff levels below.

$133

$128$140

Taipower administered industrial rate$133 /MWh

About 13 US cents per kWh, frozen through March 2026, with the state utility absorbing fuel-cost pressure that could lift it later.

These are the regulated tariff levels that define the band, not an additive cost stack. Anchored to published figures (Taipower industrial rate (Taipei Times)).

Where the cost lands

$133

Low $128Midpoint $133High $140

A band, not a single price: the cheaper end is a well-structured contract on favourable terms, the higher end a less optimised one. The midpoint is our best central read.

What is in the number: The administered Taipower industrial rate, about 13 US cents per kWh, frozen through March 2026.

Taiwan's realized cost is the state utility Taipower's administered industrial rate, so the band is tight. It is frozen for now, with the state absorbing fuel-cost pressure that could lift it later.

This market's band is free here; the ranked cross-market table with every market's precise midpoint is the Analyst Desk. See the ranked view.

Basis: triangulated from public sources, in US dollars per MWh delivered to a large high-load-factor load, excluding refundable taxes. Sources: Taipower industrial rate (Taipei Times).

Data provenance and freshness

What we hold for this market, field by field. Each tier below is assigned from the evidence actually held for that field, with provenance. A low tier means we looked and it is thin; it is not a verdict on the market.

cost: M2connection: M1reliability: M2carbon: M2regulatory: not_assessedlocal capacity: not_assessed

Against a Standardized Project Case, the binding field is cost, connection, reliability. That is what a standardized case would turn on here; a readiness result for your own project depends on your own requirements, siting and utility, which this market-level read does not know.

Read against other uses: Discovery and monitoring: M1 · Market comparison: M1 · Project Case with a carbon requirement: M1 · Conditional site diligence: insufficient.

Depth is not a rating strength, a confidence score or a feasibility conclusion. An aggregate is only ever shown for a named use case, and equals the lowest tier across that use case's required fields. Definitions v1.0. See the full coverage matrix.

How solid the data behind this rating is, on two separate axes plus recency. Source reliability is how measured the numbers are, as opposed to triangulated or modelled. Coverage is how many of the decision-relevant data dimensions (reliability, connection, and cost) are compiled for this market. The two are kept apart because they answer different questions: a market can have strong sources on the little that is covered, or broad coverage that still leans on models. Neither grades the rating itself. The rating is a cited opinion, graded separately and falsifiably by the public track record and baseline measurement plan.

Source reliability

Moderate

2 of 4 data layers measured

Coverage

Full

3 of 3 dimensions: reliability, connection, cost

Last reviewed

June 2026

per-layer vintages below

Layer	Method	Vintage	Freshness	Source
PGIQ Rating opinion	opinion	June 2026	Current	PGIQ methodology
Reliability review	modelled	2025	Fresh	on the market page
Connection friction	proxy	2025/26	Fresh	CAPRI: the energy paradox of Taiwan's AI ambition
Large-user tariff	administered	2025/26	Fresh	Taipower industrial rate (Taipei Times)
Realized cost	measured	2026	Fresh	Taipower industrial rate (Taipei Times)

Freshness reflects each layer's refresh cadence: reliability annual to three-yearly, capacity-auction and queue yearly, tariffs every year or two. The full cross-market freshness dashboard is at /data-quality.

What this market can and cannot answer yet

No market-level read closes a project case on its own, and saying only that would waste the evidence we do hold. Below is what is established, what is open, and what to do next.

What is established

Recurring cost band $128 to $140 per MWh, midpoint $133, basis administered, against the 100 MW archetype
Historical bulk-system performance (modelled)

What is open

Forward resource adequacy (assessed)
Historical utility-territory delivery performance (would change the answer)
Contractual service terms (would change the answer)
Designed site resilience
regulatory evidence
local_capacity evidence

Who would actually serve you

Not compiled for this market. Identify the serving operator before relying on any figure here, because a market-level read is an average over territories.
Who runs the connection process: not yet compiled for this market. This is wayfinding rather than evidence, and its absence is a gap in our compilation, not a fact about the market.

Credible ranges today

Cost $128 to $140 per MWh, midpoint $133 (administered)

What could disqualify this market

Supply adequacy and TSMC competition.
Geopolitics.

Five questions for the utility or system operator

What is the available capacity at the bulk supply point and substation that would serve this site, and what would a 100 MW request do to it?
Is firm service available at 100 MW, and what curtailment rights would you retain over us once we are energised?
What are the milestone dates from a complete application to full firm energisation, and which network upgrades would a 100 MW load trigger?
What is the all-in delivered rate for 100 MW at a 95% load factor, including demand, network and rider charges, under the tariff that would actually apply?
What is your own territory's SAIDI and SAIFI for the industrial class, on which major-event standard, and what redundancy serves the candidate substation?

Documents that would advance the case

The area or substation capacity study, and the current queue position list.
The applicable large-load tariff or the draft electric service agreement.
The interconnection or system impact study, and a specimen connection offer.
The current tariff sheet with all riders, and any large-load contract terms.
The utility's own reliability filing and the substation single-line diagram.