Thread continuity — Edition 8 "The next layer of value is not in grid expansion. It is in who can structurally convert financed capacity into contracted load — and who is structurally excluded from doing so." Edition 9 answers it.

The Philippines went dark in May.

From May 13–15, the NGCP issued consecutive red alerts across the Luzon and Visayas grids. Peak demand on the Luzon grid reached 12,877 MW against available capacity of 12,464 MW. Twenty-seven power plants were simultaneously offline. Transmission line trips compounded the deficit. Rotational brownouts ran up to seven hours — the first back-to-back serious shortage warnings in two years, during a declared state of national energy emergency [REPORTED — Reuters, 14 May 2026; NGCP advisories, May 2026].

The proximate cause was forced outages, transmission failures, and peak seasonal demand. The broader market context is fuel-price volatility linked to the Iran conflict — a pressure the Philippines, which imports the large majority of its crude from the Middle East, is not insulated from.

The Philippines brownouts do not prove that the absence of an intermediation layer caused the disruption. They do, however, provide a real-time illustration of the risks that emerge when commercial load remains heavily dependent on public-grid delivery during periods of simultaneous generation and transmission stress — and of the consequences when no intermediation layer exists at sufficient scale to materially buffer system-level disruption.

Edition 8 established that DFI (development finance institution) financing and utility capex are expanding grid capacity ahead of the access mechanisms required to monetise it. This edition maps who is building the intermediation layer across APAC — and what capital structure sits behind the entities positioning to fill it.


The intermediation layer is not a trading desk. It is a physical stack.

Edition 2 introduced the Effective Capacity formula:

Effective Capacity = min(IT load, power availability, cooling)

When power availability collapses — as it did across Luzon and Visayas in May — no contracted IT load or cooling infrastructure changes the outcome. The binding variable hits zero at the meter. In a market without an intermediation layer operating at sufficient scale, that is the end of the chain.

Where the intermediation layer exists, the formula behaves differently. The intermediary inserts a buffer between the transmission system and the commercial buyer — through BTM (behind-the-meter) generation, co-located firming storage, or private offtake structures that bypass the public dispatch queue. The buyer's effective capacity floor is raised. The grid's failure mode stops at the meter rather than propagating through it.

India — open access constrained by execution

Edition 8 established that India's ISTS (Inter-State Transmission System) is technically the most developed open-access transmission platform in APAC — and that its constraints define the ceiling. The execution layer has not improved.

The 100% ISTS charge waiver for solar and wind projects expired 30 June 2025. Under the CERC Fourth Amendment Regulations (notified 26 June 2025), the waiver is being phased out on a graded commissioning-date schedule: projects commissioned on or before 30 June 2025 retain a 100% waiver for 25 years; projects commissioned between July 2025 and June 2028 receive a reduced waiver, set at one of three graded bands — 75%, 50% or 25%; projects commissioned on or after 1 July 2028 receive no waiver [REPORTED — CERC Fourth Amendment Regulations, 26 June 2025; Mercom India, March 2026]. State-level surcharges compound the friction: in Maharashtra, cross-subsidy surcharges reach ₹1.69/kWh with additional surcharges of ₹1.36/kWh, rendering third-party open access commercially unviable for most industrial buyers [REPORTED — Saur Energy, May 2026].

India added 7.8 GW of solar open-access capacity in 2025, with 45+ GW in development — four times the segment's annual run-rate [REPORTED — Saur Energy, May 2026]. The pipeline is real. The access mechanism is not consistently bankable across state lines.

Australia — market-design constrained (fixed regime)

Edition 8 was precise: the CEFC (Clean Energy Finance Corporation) reduces financing cost inside the regulated NEM (National Electricity Market) framework. It does not create new access rights. Commercial aggregators work within the NEM, not around it.

What Edition 8 did not fully surface is the cost floor — and the insight it reveals. Under the National Electricity Rules, operators pay 100% of connection costs and upstream network augmentation upfront. The sector has collectively invested approximately A$3.1 billion since 2020, with a further A$7.2 billion committed to 2030 [REPORTED — Data Centres Australia, March 2026 — industry self-reported aggregate]. This cost floor changes the investment question from who can build power to who can warehouse power-access cost. That is the aggregator's structural position — and its moat.

AEMO (Australian Energy Market Operator) projects data centre consumption reaching approximately 12.0 TWh by FY2030 under the Step Change scenario, up from 3.9 TWh in FY2025 — a 25.1% average annual growth rate [REPORTED — Oxford Economics for AEMO, July 2025]. Aggregators earn value through portfolio optimisation, balancing services, hedging, demand response and structured energy procurement. As Australian market and, increasingly, regulatory expectations shift toward higher-quality delivery standards — including more time-matched physical delivery — the gap between contractual renewable positions and physical delivery becomes a growing source of both margin opportunity and risk.

Vietnam and Malaysia — limited-access / bypass-driven architecture

Edition 8 identified Vietnam's DPPA (Direct Power Purchase Agreement) framework — Decree 57/2025 — as a limited pathway to commercial access, with the routing layer increasingly shifting toward BTM and co-located structures. That dynamic is accelerating.

In May 2026, Viettel confirmed a 60 MW AI data centre near Hanoi with a dedicated power grid build — BTM architecture from a state-owned operator [REPORTED — Seoul Economic Daily, May 2026]. Vietnam's Personal Data Protection Law, effective January 2026, mandates domestic storage of Vietnamese user data — a sovereignty-driven demand signal [REPORTED].

Malaysia has signalled tighter scrutiny of non-AI data centre development and a preference for allocating scarce power and grid capacity toward higher-value AI-oriented workloads [REPORTED — PV Magazine, February 2026]. This effectively prioritises AI-related projects in the allocation queue and raises the hurdle for conventional colocation developments. The intermediation layer here is not an aggregator — it is the developer who controls both the land and the adjacent generation asset.

Philippines — no intermediation layer at scale

None of the three market structures above existed at sufficient scale in the Philippines when the grid failed in May. No ISTS equivalent. No NEM aggregator with a firming stack. No BTM developer with co-located generation insulating commercial load at scale. The grid was the only answer — and the grid was not the answer.

The Manila and Clark corridor has drawn hyperscaler evaluation interest for APAC expansion. The underwriting question is now explicit: what is the probability of a multi-day grid failure event during peak demand season, and what is the capital cost of a BTM buffer sufficient to maintain SLA (service level agreement) commitments if it occurs? That is a current project finance input — not a scenario.


1. The intermediation layer is not a single asset class — it is a market-specific capital problem.
In India, the intermediation asset is the ISTS-routed open-access PPA (Power Purchase Agreement) with state-level surcharge management — a legal and financial engineering problem as much as a generation one
In Australia, it is the NEM aggregator combining LGC (Large-scale Generation Certificate) procurement, firming BESS (Battery Energy Storage System), and demand response into a bankable offtake — a structured product problem
In Vietnam and Malaysia, it is the BTM developer controlling land, generation, and grid connection simultaneously — a real assets and permitting problem
Each market requires a different capital structure and risk allocation to win the intermediation position
A pan-APAC intermediation fund applying one structure across all three markets will misprice risk in at least two of them
2. Australia's connection cost floor changes the investment question from who can build power to who can warehouse power-access cost.
Operators have collectively paid A$3.1 billion since 2020, with A$7.2 billion committed to 2030 — all under the NEM's full-cost-recovery rules [REPORTED — Data Centres Australia, March 2026 — industry self-reported aggregate]
Mid-market buyers who cannot absorb upfront connection costs at hyperscaler scale are structurally dependent on aggregators who can warehouse that cost and amortise it across a portfolio
The aggregator's moat is not generation procurement — it is balance sheet capacity to absorb the cost floor, combined with LGC and BESS assembly that mid-market buyers cannot self-replicate
AEMO projects data centre consumption reaching 12.0 TWh by FY2030 at 25.1% annual growth [REPORTED — Oxford Economics for AEMO, July 2025] — the aggregator's addressable market is compounding in real time
As regulatory expectations move toward time-matched physical delivery, the gap between LGC contractual positions and physical delivery becomes an additional source of both value and risk for aggregators
3. India's ISTS charge phase-out is redrawing the access economics for the entire 45+ GW open-access pipeline.
The waiver steps down across three graded bands — 75%, 50% or 25% — for projects commissioned between July 2025 and June 2028, with no waiver for projects commissioned on or after 1 July 2028 [REPORTED — CERC Fourth Amendment Regulations, 26 June 2025; Mercom India, March 2026]
Developers locking in intra-state procurement structures in high-irradiation states before the waiver cutoff of 1 July 2028 hold a structural cost advantage
State-level surcharge divergence — Maharashtra at ₹3.05/kWh combined versus Gujarat's more favourable regime — creates a geography-of-access problem invisible in national pipeline figures
The intermediation moat in India is state-level access structuring solved in two or three key industrial states — not generation scale
4. Philippines: BTM generation and BESS are increasingly becoming a prerequisite for new hyperscale-grade data centre underwriting in-market.
The May events make the BTM buffer question a current underwriting input, not a scenario, for the Manila and Clark corridor
First mover in co-located or campus-adjacent generation with grid-islanding capability holds a position that cannot be replicated quickly given permitting and procurement lead times
BTM + BESS is increasingly becoming a prerequisite for new hyperscale-grade data centre underwriting in-market — not a premium option
Investment Lens · Edition 9
ConstraintIndia — ISTS charge phase-out
ImpactInter-state open access economics deteriorating; intra-state structures now preferred
Capital responseState-specific access structuring; group-captive PPA in Gujarat and Karnataka pre-2028
Winning assetC&I developer with intra-state access solved in 2–3 key industrial states before the 1 July 2028 waiver cutoff
ConstraintAustralia — NEM connection cost floor
ImpactMid-market buyers cannot self-fund; structurally aggregator-dependent
Capital responseNEM aggregator warehousing connection cost + bundling LGC procurement and firming BESS
Winning assetAggregator with balance sheet capacity and multi-offtaker portfolio; not a single-tenant structure
ConstraintVietnam / Malaysia — grid access uncertainty
ImpactHyperscale load bypassing public grid; BTM architecture is default, not premium
Capital responseBTM developer controlling land + adjacent generation + connection; co-located structure
Winning assetDeveloper with land bank adjacent to industrial zones and permitted generation capacity
ConstraintPhilippines — absent intermediation layer at scale
ImpactGrid failure propagates to commercial load; no buffer operating at sufficient scale
Capital responseBTM generation + BESS increasingly becoming a prerequisite for new hyperscale-grade data centre underwriting
Winning assetFirst mover in co-located or campus-adjacent generation with grid-islanding capability

The intermediation gap is not the same problem in every APAC market. In India it is a regulatory navigation and surcharge-structuring problem. In Australia it is a balance sheet and structured product problem — with a growing delivery-quality dimension as market and, increasingly, regulatory expectations shift toward more time-matched physical delivery. In Vietnam and Malaysia it is a real assets and permitting problem. In the Philippines it is an absence-at-scale problem — and the May events showed the cost of that absence in operational terms.

What connects all four is the thesis Edition 8 established: DFI financing and regulated utility capex are expanding grid infrastructure faster than the access mechanisms required to monetise it. The transmission investment is happening. The contracts between that investment and commercial buyers are not. The intermediation layer is where that gap closes.

The capital structures that close it look materially different across each market.

Risk flag

The intermediation thesis ultimately rests on the credit quality of the anchor tenants underwriting long-duration offtake commitments. As hyperscalers increasingly finance expansion through debt markets, tighter credit conditions could affect project-level underwriting assumptions even where power-access fundamentals remain supportive. Capital structures for intermediation vehicles should not be sized to a single anchor tenant.

The question for Edition 10 is whether the entities currently positioning in each market are building intermediation as a standalone business — or as a defensive moat for an adjacent position they already hold. That distinction will determine whether the intermediation layer remains a specialist lane or gets absorbed into the vertical stacks of the operators who need it most.
Edition 10 picks that up.