Hong Kong: PGIQ Rating, Tier 4 Constrained

The same market, read four ways

BuildSellPlanMonitor

Hong Kong is a strong market to sell power into

Grade A (Strong)

Offtake score 87/100

Demand 53, build feasibility 65, 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 $188/MWh delivered here. A competitive supply offer works against that level.

Connection

Connection difficulty Moderate, typical wait 2-3 years.

See it on the Offtake Grade →

Hong Kong: room to spare at system level, light demand pressure

Room to spare 72/100

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 (Room to spare), and how hard demand is already pushing (Light). 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

Sound

Manageable with normal diligence · grid strength 67/100

A cited, dated grid-risk read on Hong Kong 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-High confidence. Scored on a public, dated, falsifiable record you can cite in a credit memo.

Delivery reliability

Reviewed; no delivery reliability flags found in the bulk and local outage record.

Future firm supply

Reviewed; firm supply ample.

Cost certainty

Recurring cost about $188/MWh (band $170 to $210) for the standard 100 MW case, basis administered. Excludes connection capital.

Regulatory / stroke-of-pen

Not assessed.

Execution risk

Connection friction Moderate, 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: CLP and HK Electric 2026 tariffs (HK Government) · CLP supply reliability about 99.999% · Developing data centres in Hong Kong (HK Government)

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%42

Constrained access: severe land scarcity and a small system, though the government facilitates data centres and is upgrading the Northern Metropolis grid for AI density.

Availability25%55

Strong local delivery record: CLP reports about 99.999% supply reliability, roughly 2.6 minutes of unplanned outage a year, on a ~7.2 GW system.

Cost25%25

Very high cost: regulated Scheme-of-Control net tariffs run about 141 to 163 HK cents per kWh (~$180 to $210 per MWh).

Momentum15%58

Moderate momentum: an established banking and cloud hub plus a government AI Supercomputing Centre, but a small overall market.

Carbon5%40

Moderate-high carbon: gas and coal, with imported nuclear from Daya Bay.

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 reviewed: no material concerns were found, so the rating is unchanged. 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: No reliability flags found

Historical utility-territory delivery performance: No reliability flags found

Strong local delivery record on the utility's own published measure: CLP reports about 99.999% supply reliability, roughly 2.6 minutes of unplanned outage a year, on a small, dense system. This describes distribution performance, not the terms of a firm service contract.

measured · CLP supply reliability and power quality

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: Med-High. 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

Moderate

Typical connection wait

2-3 years

Source: Hong Kong data-centre power supply (government)

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.

CLP and HK Electric supply an orderly but land- and power-constrained market adding about 400 MW by 2026.

Hong Kong's two vertically integrated utilities, CLP and HK Electric, run an orderly connection process, and the city already hosts about 1.42 GW of data-centre IT load with more than 400 MW more due by end-2026. The constraint is land and long-run generation capacity in a dense city, not a chaotic queue.

What builds the 30-month wait

Application & CLP / HK Electric study6 mo

Supply agreement6 mo

Network works12 mo

Construction & energization6 mo

Typical time to energization30 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 (Hong Kong data-centre power supply (government)).

Data-center share of demand pressure: High. Sources: Hong Kong data-centre power supply (government) · CLP apply for grid connection. The full cross-market connection dataset is available through the Connection Friction Feed.

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

$170 to $210 per MWh

Midpoint about $188 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.

$180

$209

$170$210

CLP Power net tariff (Kowloon & New Territories)$180 /MWh

About 140.6 HK cents per kWh for 2026 under the Scheme of Control, roughly $180 per MWh. CLP serves most of Hong Kong, including the data-center clusters.

HK Electric net tariff (Hong Kong Island)$209 /MWh

About 163.3 HK cents per kWh for 2026, roughly $209 per MWh, the pricier of the two regulated utilities.

These are the regulated tariff levels that define the band, not an additive cost stack. Hong Kong has no wholesale market; two vertically integrated utilities set regulated Scheme-of-Control tariffs, so the band spans the two providers. Anchored to published figures (CLP and HK Electric 2026 tariffs (HK Government)).

Where the cost lands

$188

Low $170Midpoint $188High $210

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: Regulated Scheme-of-Control net tariffs from the two utilities: CLP Power at about 140.6 HK cents per kWh and HK Electric at about 163.3 HK cents per kWh for 2026, roughly $180 to $210 per MWh. There is no wholesale market.

Hong Kong's realized cost is the regulated Scheme-of-Control tariff of whichever of the two utilities serves the site, among the highest in the region. There is no wholesale market, so the tariff is the delivered cost.

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: CLP and HK Electric 2026 tariffs (HK Government) · CLP 2026 tariff tables.

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: M1carbon: 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 3 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	measured	2025	Fresh	CLP supply reliability and power quality
Connection friction	proxy	2025/26	Fresh	Hong Kong data-centre power supply (government)
Realized cost	measured	2026	Fresh	CLP and HK Electric 2026 tariffs (HK Government)

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 $170 to $210 per MWh, midpoint $188, basis administered, against the 100 MW archetype

What is open

Historical bulk-system performance
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 $170 to $210 per MWh, midpoint $188 (administered)

What could disqualify this market

Very high power cost.
Land and space scarcity on a small system.

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.

Evidence behind this rating

The real-world events that test our Tier 4 call for Hong Kong: 2 support it. Each is dated and cited, so you can check the rating against what is actually happening on the ground.

Hong Kong utilities report about 99.999% supply reliability as data-centre demand grows ✓ Supports the rating

2025AvailabilityMateriality: med

CLP reports about 99.999% supply reliability, roughly 2.6 minutes of unplanned outage a year, and both utilities say they hold adequate capacity and backup. That is a strong local delivery record on the utility's own published measure, and a real strength for a large load, even as high cost and land scarcity constrain the rating. It describes distribution performance, not the terms of a firm service contract.

Source: CLP ↗

Government builds a 3,000-petaflop AI Supercomputing Centre and upgrades the Northern Metropolis grid ✓ Supports the rating

2025MomentumMateriality: low

Hong Kong is standing up a government AI Supercomputing Centre and CLP is upgrading the Northern Metropolis grid for kilowatt-density AI load, real but modest momentum on a small, land-constrained system.

Source: Developing data centres in Hong Kong (HK Government) ↗

Cited, dated developments tagged to the pillar they bear on, and whether they support or challenge the current tier. A material item flags the market for analyst review and may drive a rating action.