In the last Bond Voyage edition of 2025, we examine the vast amounts of capital being deployed to build new high-capacity data centres and how the expenditure is being funded in the bond markets.
Read this article to understand:
- How the “growth” story for artificial intelligence (AI) has changed to one of “cash flows”
- The vast amounts of capital being invested in new data centre infrastructure
- The impact of this capital raising on the bond markets
US hyperscalers (large cloud service providers) are spending vast sums to build high-capacity data centres, aiming to keep pace with generative AI technology. These investments are no longer optional – they are essential for competitiveness – and they are beginning to influence bond markets.
Investment in generative AI is expected to expand considerably
Generative AI has made data centres a focal point for global investment. It is seen as a potentially game-changing technology, and companies are committing extraordinary amounts of capital to avoid falling behind. From semiconductor and chip makers to hyperscalers, the scale of investment is staggering.
McKinsey estimates nearly $7 trillion will be required by 2030 to meet the growing demand for computing power – both for AI and traditional needs.1 While chip makers’ ability to monetise this trend is clearer, questions remain about the size of future economic benefits for others. What is certain: the investment wave has begun, and its impact on credit markets is already visible.
Hyperscalers in particular are spending billions on data centre infrastructure (see Figure 1). This capital expenditure is increasingly viewed not as discretionary, but as critical to maintaining competitiveness.
Figure 1: Expanding capital expenditure across hyperscalers ($ billion)
Note: Annual data. Reported figures for 2022-2024, forecasts for 2025 onwards.
Source: Aviva Investors, Bloomberg. Data as of December 3, 2025.
Investment-grade bond market as a funding source
Hyperscalers have traditionally enjoyed strong investment-grade ratings and robust cash flows, meaning they rarely needed to tap bond markets. Leverage has generally been low, and when they did issue debt, it was primarily for cost-efficient funding.
The scale of current capital expenditure is changing this picture. Companies are now issuing bonds more frequently to finance these projects, even as they continue to prioritise shareholder returns. While a significant portion of spending will still come from existing cash reserves, reliance on bond markets is set to grow.
- Top-tier issuers: Highly-rated hyperscalers can raise debt on attractive terms, given their low cost of debt relative to equity. They retain flexibility to reduce issuance if market conditions deteriorate.
- Lower-rated issuers: May have little choice but to rely heavily on debt issuance, as free cash flow alone cannot cover the required investment.
The expectation of substantial new issuance has weighed on sentiment, sparking concerns about a potential “Gen AI bubble”. This shift has also changed the narrative: discussions that once focused on productivity and growth – fuelling equity rallies – are now centred on cash flows and credit markets.
Data centre lessors also seek funding
There is another layer to the data centre investment story. Many hyperscalers lease the physical “shell” – the land and buildings – rather than owning them outright. The developers of these facilities are also turning to capital markets for funding.
The ability of hyperscalers to pay rent is a key factor when assessing these structures
The creditworthiness of these developers’ bonds is closely linked to the financial strength of their tenants. The ability of hyperscalers to pay rent is therefore a key factor when assessing these structures.
This has led to the creation of new issuance models where tenants provide implicit credit support to lessors without directly guaranteeing the debt. These arrangements give investors confidence by linking the bonds to the tenants’ strong credit quality.
However, this approach introduces additional concentration risk. As cash flows and creditworthiness become more interconnected across legally separate entities, investors must remain alert to indirect exposures when evaluating these structures.
Impact on bond markets
This is not just an investment-grade story – some high-yield issuers are also involved. However, the investment-grade market is much larger and better positioned to absorb the expected surge in AI-related issuance.
The potential volume of new investment-grade bonds could significantly increase the technology sector’s share of credit benchmarks. Currently, technology accounts for 6.7 per cent of the index, with around $910 billion outstanding.2 J.P. Morgan estimates that $300 billion of AI/data centre-related bonds may be issued over the next year, lifting technology’s share to 8.7 per cent.
If the anticipated $1.5 trillion of debt is eventually issued, technology could represent as much as 15.9 per cent of the index. For credit investors, the key challenge will be ensuring issuers’ balance sheets remain resilient amid rising capital expenditure and leverage.
As hyperscalers’ data centre leases continue to grow, traditional debt metrics may not tell the full story. We expect lease-adjusted debt to become an increasingly important measure when assessing balance sheets, given the potential scale of these commitments.
