Edition 24 ended with a question: "So how much of the value created upstream ultimately survives into the owner's return?" This edition takes that question into the credit markets.
Reuters reported that AI-related issuance by low-rated firms has reached roughly US$88 billion so far in 2026, with most borrowing coming from US issuers — compared with about US$20 billion in the first eleven months of 2025. [source] That is a financing signal: AI infrastructure is increasingly being tested in credit markets, where investors have to distinguish between projects that can and cannot support external capital.
The scarce megawatt is not necessarily the valuable megawatt.
Neither is the first megawatt to secure land, obtain a grid position, or receive an allocation. The question is what happens next — when does physical scarcity become cash flow that an investor can actually underwrite?
Data-centre scarcity is often discussed as though there were one bottleneck. There isn't. Land can be scarce. Grid access can be scarce. Power allocation can be scarce. And a credible delivery window can itself become scarce. Each layer can carry option value.
Keppel's planned AI data-centre campus near Morwell, Victoria, shows how that value can begin well before a facility exists. Keppel secured rights to lease a 123-hectare site with up to 720 MW of gross power capacity, and pays an annual access fee while it undertakes pre-development work, including planning approvals and contracting power and water. [source]
Transgrid says Sydney currently has insufficient transmission capacity to connect any of the 20 GW of new large-load connection enquiries it has received. Additional data-centre connections depend on new user-funded transmission upgrades. [source] In Singapore, the government provisionally allocated 200 MW of new data-centre capacity to four operators under DC-CFA2. [source]
Land rights, grid positions and regulatory allocations create option value. They tell us that someone controls an opportunity. They do not yet tell us whether that opportunity will become an underwritable asset.
AEMO reported 11 large data-centre projects representing 5.4 GW of maximum demand in the transmission connection process in Q1 2026 — 4.1 GW at application stage and 1.3 GW at proponent implementation stage. [source] That is evidence of progress toward deliverability. It is not evidence that those MW have reached it.
The same MW label covers at least five structurally different states. They are not a funnel — they do not describe one dataset shrinking mechanically from top to bottom. They describe different risks being removed, each answering a different underwriting question.
Malaysia's TNB Green Lane has compressed the electricity-supply implementation process from 36–48 months to as little as 12 months. By end-2024, TNB reported 18 completed supply connections representing 1,856 MW; during 2024, it signed 37 ESAs representing 5,765 MW. [source]
Power contract ≠ customer contract.
A signed ESA evidences a utility-side supply commitment or pathway. It does not confirm a tenant/customer contract. TNB's completed connections provide the stronger evidence of delivered electricity supply. That distinction is what separates Deliverable MW from Revenue MW.
Keppel DC REIT's 1H 2026 results report portfolio contracted power capacity of approximately 95%, which it defines as the proportion of revenue-generating power capacity, alongside contract renewals in Singapore and Australia and a WALE of 6.7 years by lettable area. [source] At Revenue MW, the capacity is no longer merely available — someone is paying for it.
Financeable MW is where contract, counterparty, delivery, collateral and risk allocation together support external capital. Five gates need to align sufficiently for external capital to underwrite the project:
Counterparty credit quality
Contract durability
Physical delivery credibility
Collateral quality
Risk allocation between owner, lender and tenant
Weakness in any one gate can materially change the financing structure, pricing, recourse or amount of external capital available. Financing structure and credit support — senior secured debt, non-recourse SPVs, guarantees and other protections — are shaped by how those gates are resolved; they are not themselves a gate.
The examples above show why headline pipeline depth can overstate the amount of capacity that has actually reached the financeable state.
Hut 8's Beacon Point project in Texas provides a real-world test of Financeable MW.
In June, Beacon Point DC issued US$4.25 billion of 6.129% senior secured notes due November 2042. The proceeds are financing a data centre with 352 MW of critical IT capacity and associated substation infrastructure. [source]
The facility is leased in its entirety to a high-investment-grade company under a triple-net lease. Hut 8's filing states that rent under that lease is expected to be the principal source of cash flow servicing the notes.
The debt is secured by first-priority liens over substantially all of Beacon Point DC's assets and by a pledge of equity interests in the project company. The obligations are not guaranteed by Hut 8 or the tenant.
→ long-term customer contract
→ high-investment-grade counterparty
→ contracted rent as principal debt-service source
→ first-priority project liens
→ non-recourse SPV structure
Project Jupiter shows why Financeable MW is not necessarily a permanent state. Oracle recently issued a force-majeure notice relating to its New Mexico campus following setbacks in its power-delivery pathway. The notice reportedly preserves Oracle's ability to delay the start of higher rental payments if the project misses its planned 2028 opening, although Oracle says the project remains on schedule and Blue Owl says the notice does not change the project's financial commitments. [source]
The underwriting point is narrower: a signed tenant contract does not eliminate delivery risk. If the power pathway slips, contractual protections can activate, rent timing can change, and the economics supporting external capital may need to be reassessed.
Financeable MW therefore has to be monitored, not merely achieved.
Meta holds a 20% membership interest in a Louisiana data-centre joint venture. The parties have committed to fund their respective pro-rata shares of approximately US$27 billion of total estimated development costs. [source]
Meta also disclosed approximately US$12.31 billion of initial lease commitments, residual-value guarantees with an aggregate threshold of approximately US$28 billion, and maximum exposure to loss of approximately US$46.03 billion as of June 2026 (a maximum-exposure measure, not an estimate of expected loss).
Separately, Meta states that residual-value guarantee payments were assessed as not probable and no RVG liability had been recorded.
Minority ownership is not minority economic exposure. And moving the financing structure is not the same as moving all the risk.
A pipeline figure that does not specify MW state conflates five structurally different levels of investment readiness.
Nominal and Controlled MW may carry option value through site rights, grid position and development optionality. They are not yet cash flow.
The transition from Revenue MW to Financeable MW requires structural work distinct from building the data centre: contract structure, counterparty selection, collateral packaging and risk allocation between owner, lender and tenant.
Beacon Point illustrates one form of project financeability. The diligence question for markets such as Johor is whether contractual and risk-allocation structures are developing alongside the MW pipeline.
The last few editions have moved from where a megawatt needs to be, to when it can arrive, to what state it has actually reached.
Delivery creates usable capacity.
A bankable contract turns that capacity into financeable value.
The MW-state framework can be applied across markets and scales. Transgrid's 20 GW is at Nominal. Keppel Morwell and Singapore's DC-CFA2 200 MW are at Controlled. TNB Green Lane is moving capacity toward Deliverable. Keppel DC REIT's 95% contracted power capacity demonstrates Revenue. Beacon Point has reached Financeable.
Each state can carry value. But Financeable MW is the state where the project's ability to support external capital has been explicitly demonstrated.
The next question shifts from infrastructure to counterparty.
What changes when the counterparty behind the bankable contract is a neocloud rather than a hyperscaler?
Analysis relies primarily on regulatory filings, official operator and government disclosures, supplemented by Reuters market reporting. Sources current as of 1 October 2026.
Glossary — acronyms used in this edition
| Acronym | Full name | Plain English |
|---|---|---|
| MW | Megawatt | Unit of electrical power (1,000 kW) |
| GW | Gigawatt | 1,000 MW |
| DC-CFA2 | Data Centre Call for Applications 2 | Singapore's second competitive allocation process for new data-centre capacity |
| AEMO | Australian Energy Market Operator | Operator of Australia's major electricity and gas markets and systems |
| TNB | Tenaga Nasional Berhad | Malaysia's national electricity utility |
| ESA | Electricity Supply Agreement | Utility-side electricity supply arrangement |
| NNN | Triple-net lease | Lease structure in which specified operating costs are borne by the tenant |
| SPV | Special-purpose vehicle | Ring-fenced legal entity used to hold/project-finance an asset |
| WALE | Weighted Average Lease Expiry | Average remaining lease duration across a portfolio |
| RVG | Residual Value Guarantee | Contractual protection tied to an asset's future residual value |
| DSCR | Debt Service Coverage Ratio | Project cash flow relative to debt obligations; core project-finance underwriting metric |