TL;DR
Asia Private Equity Review tracked US$32.4 billion of H1 2026 credit and loans secured by PE portfolio companies; data centres represented 46% of its industry split. That is a concentration signal to investigate, not a measure of the whole Asia-Pacific private-credit market.
The headline number needs its denominator
Asia Private Equity Review’s H1 2026 data brief records US$32.4 billion of “private credit and loans secured by PE portfolio companies”. Its industry chart assigns 46% to data centres, ahead of services and healthcare at 10% each, industrial goods at 9% and energy and natural resources at 7%. A separate market split attributes 47% to Australia and New Zealand and 13% to Southeast Asia.
Those figures make data-centre financing an important diligence subject. They do not show that 46% of all Asia-Pacific private credit is exposed to the sector. The chart covers a defined set of credit and loans linked to private-equity portfolio companies. It does not publish the number of borrowers, the concentration by sponsor, the size of individual facilities, loan seniority, currencies, tenors or whether commitments were fully drawn.
That distinction is not pedantic. A few very large facilities can dominate a transaction-value dataset without describing the average loan, a diversified fund or the asset class as a whole. An allocator should therefore read 46% as a prompt to request the underlying exposure schedule, not as a market-allocation target.
Real financing demand sits behind the statistic
The capital requirement is not hypothetical. Mingtiandi reported in March that Singapore-based Princeton Digital Group planned to raise up to US$5 billion of debt in 2026 for its Asia-Pacific hyperscale pipeline. It had expanded an existing facility to US$750 million with a consortium of global banks. That transaction is bank financing, not evidence that every data-centre loan belongs in a private-credit portfolio, but it demonstrates the scale of funding that platform expansion can require.
A current 17 July analysis by The Asset describes AI infrastructure as a mainstream institutional investment opportunity. The investable question, however, is not whether AI increases demand for computing. It is whether each borrower can turn contracted capacity into cash flow before construction, power and refinancing risks consume the expected credit margin.
1. How many exposures sit behind 46%?
Start with the loan tape. Ask for unique borrowers, sponsors, facilities and countries; committed and drawn amounts; origination dates; loan-to-cost and loan-to-value measures; and each position’s share of fund net asset value. A manager should also identify exposure duplicated through co-investments, parallel funds, warehouses or managed accounts.
The geography needs the same discipline. Australia and New Zealand plus Southeast Asia account for 60% of the dataset’s market split, but the published chart does not cross-tabulate market and industry. It cannot establish that the data-centre share is concentrated in those markets. Only the underlying records can answer that question.
2. When does contracted demand become lender cash flow?
Data centres can have attractive long-term contracts, but “contracted” is not a complete credit description. Allocators need the tenant, parent guarantee, termination rights, pricing escalators, commencement conditions and responsibility for fit-out. They should distinguish a fully operational facility from land, powered shell, construction work in progress and capacity reserved before completion.
Customer concentration deserves particular attention. One highly rated tenant may strengthen a facility’s cash flow, yet a bespoke design can reduce reletting flexibility if that tenant leaves. The underwriting case should show how revenue, debt service and recovery value behave under a delayed start, slower ramp-up, tenant downgrade or non-renewal.
3. Are power and construction milestones financeable?
Goldman Sachs identified power as a primary constraint on data-centre construction and delivery in Asia. Its June outlook also noted that AI customers can commit before construction and require different cooling and rack-density designs. Those conditions can support pre-leasing, but they can also introduce redesign, equipment, grid-connection and completion risk.
Credit papers should therefore specify secured power capacity, connection dates, tariff assumptions, water and cooling requirements, construction contingencies and completion support. Covenants should be linked to milestones that matter to cash generation, not merely to headline megawatts. Allocators should ask who supplies extra equity if commissioning slips or equipment costs rise.
4. Which debt layer is being underwritten?
The primary dataset combines private credit with loans secured by PE portfolio companies. That makes capital-structure classification essential. Senior project debt, asset-backed facilities, holding-company loans, acquisition finance and preferred capital can all finance the same platform while offering very different collateral, covenants and recovery prospects.
For each exposure, investors need security ranking, intercreditor terms, maintenance covenants, cash sweeps, permitted additional debt, sponsor support, hedging, maturity and refinancing assumptions. They should also test whether an ostensibly diversified portfolio ultimately depends on the same hyperscalers, utilities, equipment suppliers or exit market.
What the statistic can and cannot do
The 46% share usefully shows that data centres were the largest industry in this H1 dataset. It does not establish market-wide exposure, superior returns, low defaults or sufficient diversification. Nor does it show whether lenders were paid adequately for construction, tenant, power and refinancing risks.
The right allocator response is an evidence request: reconcile the chart to borrower-level positions, separate operating assets from development risk, map common counterparties and stress the maturity schedule. If a manager cannot produce that bridge, the problem is not the data-centre theme. It is the opacity of the underwriting.
Source note: The US$32.4 billion total and percentage splits come from Asia Private Equity Centre’s H1 2026 data brief, published on 16 July 2026. Independent market context was checked against The Asset, Goldman Sachs and Mingtiandi. This is market analysis, not investment advice. No return, default or allocation forecast is made.
Frequently Asked Questions
What does the 46% data-centre figure measure?
It is the data-centre share of the industry split in Asia Private Equity Review’s H1 2026 dataset covering US$32.4 billion of private credit and loans secured by PE portfolio companies.
Does it mean 46% of all Asia-Pacific private credit is in data centres?
No. The chart describes one PE portfolio-company credit dataset. It is not presented as the asset-class-wide share of Asia-Pacific private-credit AUM, lending or investor exposure.
Why does power matter in data-centre credit underwriting?
A delayed grid connection or insufficient power can postpone commissioning and revenue. Lenders therefore need evidence on capacity, connection milestones, tariffs, backup arrangements and who bears delay or cost-overrun risk.
What should an allocator ask a manager first?
Ask for the underlying denominator: unique borrowers, sponsors, countries, facility types, seniority, committed and drawn amounts, tenant concentration and overlap across funds or managed accounts.