Key facts
- Investors are growing skeptical of the AI infrastructure buildout.
- The BIS warns that financing for the AI infrastructure boom is opaque and debt-heavy.
- Hyperscalers are set to spend over US$1 trillion on AI-related capital expenditure in 2025 and 2026.
- Private credit funds provided over US$40 billion in loans to AI-related companies in 2025.
- Malaysia's Johor region attracted approximately US$39 billion in data center investment by Q2 2025.
- Malaysia has paused approvals for non-AI facilities due to power and water constraints.
Data center companies across Asia are experiencing increased investor scrutiny due to a series of stalled initial public offerings and volatile stock market debuts. This trend raises questions about the valuations within the booming AI infrastructure business.
The Bank for International Settlements (BIS) has highlighted the fragility of the financing supporting the global AI infrastructure boom, describing it as opaque and heavily reliant on debt. The BIS noted that hyperscalers are set to spend over US$1 trillion on AI-related capital expenditure in 2025 and 2026, a pace that is outstripping their earnings, leading them to increase borrowing.
Private credit funds have become a significant source of funding, originating over US$40 billion in loans to AI-related companies in 2025. The BIS warns of a web of circular financing and poorly disclosed terms, increasing the risk of assets being pledged multiple times. Asia, having absorbed significant demand spillover, is particularly exposed.
Malaysia's Johor region has attracted substantial data center investment, with approximately US$39 billion committed by the second quarter of 2025. However, the country's government has paused approvals for non-AI facilities due to constraints on power and water resources, indicating potential supply-side roadblocks.