Two Tracks, No Common Benchmark — how APAC is pricing data centre cash flows through formats whose prices do not translate directly

In July 2025, the institutional order book for a Chinese data centre REIT was 166 times covered at the final offering price. The retail tranche was 456 times over-subscribed, and the offering closed ahead of schedule.

That transaction — GDS Holdings' C-REIT, listed on the Shanghai Stock Exchange under fund code 508060, raising RMB 2.4 billion seeded with a 29MW facility near Shanghai — was one of two related but separate capital-recycling transactions GDS completed in China in 2025. In March, it sold project companies into a private REIT financed through an asset-backed security, retaining 30% of the ABS and ongoing operating rights, with Chinese institutions led by China Life subscribing the balance. In July, a separate public C-REIT raised RMB 2.4 billion to acquire a different project company, holding a 29MW data centre project in Kunshan, Jiangsu, bordering Shanghai. GDS reinvested RMB 480 million to retain 20%.

The March private REIT was designed to permit eventual injection into a public C-REIT once qualified. The July C-REIT was not that injection. It was a separate asset monetisation, with separate assets, separate proceeds, and separate accounting: a deconsolidation gain of approximately RMB 1,057 million recognised in Q1 2025 for the ABS transaction, and RMB 1,369.3 million in Q3 2025 for the project companies sold to the C-REIT (GDS Holdings).

Meanwhile, in the international lease-backed term ABS market, the regional template was still unformed in the latest sources cited here. IFR reported in March 2026 that market participants in Asia were still waiting for a first lease-backed term data centre ABS to establish an international-format template, and RBC Capital Markets reported in January 2026 that Australia was expected to launch the region's first such transaction during 2026, with panel participants projecting A$2–7 billion in annual issuance at maturity. Whether that position still held at publication could not be independently confirmed from open sources.

Edition 15 asked what has to be true about the tenant, or the building, for the debt above it to survive a tenant that does not renew. The answer turns out to depend on which market you are standing in.

This edition uses securitisation in two senses. Broadly, both structures convert data centre cash flows into investable securities. Narrowly, the benchmark discussed here refers to lease-backed term debt priced through spreads, leverage, and repayment protection. Ordinary listed REITs and listed data centre operators may provide valuation comparables, but they are not direct credit-spread benchmarks.

The distinction is not geographic. It is structural, and the difference determines who bears what.

Why the two prices do not translate directly

Track A prices credit risk: will contracted rent service senior debt through a repayment date. Track B prices long-dated fund-unit risk: what is a claim on project-company cash flow, over a 38-year fund term, worth today. A 166x institutional order book is not a tight spread — it is a C-REIT unit subscription ratio, and the two quantities are not convertible. Investors differ accordingly: Chinese insurance capital and domestic retail on one side; institutional credit buyers, domestic and offshore, on the other.

What makes Track A structurally harder in APAC

Adoption faces fragmented markets and different legal systems across the region (IFR, March 2026). APAC remains early in its build cycle, with many facilities under construction and few mature candidates suitable for ABS financing. Roughly US$116 billion may be required to build out the existing APAC colocation pipeline over five to seven years, excluding hyperscale projects (Cushman & Wakefield, February 2025). Capital recycling through REITs already exists in markets including Singapore and Australia, but generally requires selling down ownership or control — which is precisely the trade Track A is designed to avoid.

Edition 14 separated announced megawatts from bankable megawatts. Edition 15 followed the financing-structure gate into the years when leases renew and debt must be repaid. Edition 16 finds that APAC is answering that question in two different languages at once.

The constraint has moved again, and not in the direction the previous editions implied. It is not that APAC cannot price data centre cash flow — China has done it, at demand levels that would be remarkable in any market. It is that there is no common APAC benchmark, because there is no single APAC instrument.

The practical consequence: a Chinese subscription ratio tells you about demand for scarce listed product; an Australian lease-backed ABS spread primarily tells you about credit pricing. Neither translates directly into the other. The temptation to read one as evidence about the other will be strongest precisely when a fiduciary most wants a regional comparable — at the moment of a debut, with no domestic precedent to reference.

The two tracks may become more internally benchmarkable without converging. As each format accumulates repeat transactions, investors may gain better comparables within that structure. But a Chinese C-REIT unit and an international lease-backed term ABS would still price different claims, transfer different risks, and serve different investor bases.

Scarcity migrates, capital follows the new bottleneck, and asset values reprice around it. For a regional allocator, the immediate bottleneck is not the absence of capital or transactions. It is comparability.

If a Chinese C-REIT unit and an Australian lease-backed ABS price different claims on the same asset class, what should a regional allocator underwrite against when an APAC data centre CMBS is next brought to market?