Union Connect: How Hyperscalers Are Reshaping Global Bond Markets
|
|
|
|
||||
|
|
|
|
Hyperscalers are investing heavily in AI infrastructure and are becoming increasingly important issuers in global credit markets. In this interview, Annika Frilling and Moritz Krohn discuss the implications for bond investors, key credit considerations and the longer-term impact on market structures.
Moritz, few topics are currently attracting as much attention from institutional investors as the bond issuance of the major hyperscalers. Why is this segment so firmly in focus at present?
Moritz Krohn: The momentum is exceptional. Companies such as Microsoft, Amazon, Alphabet, Meta and Oracle are investing billions in AI infrastructure, particularly in data centres, cloud capacity and high-performance chips.
From a credit investor's perspective, the key question is how these investments are being financed and what impact they will have on the companies' credit profiles. At the same time, we are seeing hyperscalers become an increasingly important component of the global corporate bond market. Investors are therefore primarily asking three questions: Does the business model remain attractive? Can the investments be monetised on a sustainable basis? And how are these issuers changing the structure of credit markets?
From a credit investor's perspective, the key question is how these investments are being financed and what impact they will have on the companies' credit profiles. At the same time, we are seeing hyperscalers become an increasingly important component of the global corporate bond market. Investors are therefore primarily asking three questions: Does the business model remain attractive? Can the investments be monetised on a sustainable basis? And how are these issuers changing the structure of credit markets?
Let us start with the business model. Does the investment case for hyperscalers remain intact despite the enormous level of investment?
Moritz Krohn: From our perspective, the answer is broadly yes. The use of AI applications continues to grow at a very dynamic pace. As a result, demand for computing power, storage and cloud capacity is increasing.
We still see a structural excess demand for AI infrastructure at present. For providers, this means that the key challenge is not a lack of demand, but the ability to provide sufficient capacity.
This is important for credit investors because a robust demand environment forms the basis for justifying the high levels of investment through rising future cash flows and earnings.
We still see a structural excess demand for AI infrastructure at present. For providers, this means that the key challenge is not a lack of demand, but the ability to provide sufficient capacity.
This is important for credit investors because a robust demand environment forms the basis for justifying the high levels of investment through rising future cash flows and earnings.
At the same time, powerful Chinese AI models are becoming increasingly important. Does this change the perspective on hyperscalers?
Moritz Krohn: This development is often misunderstood. Open-source or open-weight does not automatically mean free of charge. Many of the new Chinese models are also monetised through commercial offerings and compete primarily on the basis of an attractive price-performance ratio.
For the major cloud providers, greater model diversity may even be a positive factor. Microsoft, Amazon and Alphabet see themselves as multi-model platforms. What matters most to them is not which individual model dominates the market, but that the overall use of AI continues to increase, thereby driving higher utilisation of their infrastructure.
For the major cloud providers, greater model diversity may even be a positive factor. Microsoft, Amazon and Alphabet see themselves as multi-model platforms. What matters most to them is not which individual model dominates the market, but that the overall use of AI continues to increase, thereby driving higher utilisation of their infrastructure.
Critics are warning of overinvestment in AI infrastructure. How do you assess this risk?
Moritz Krohn: This is probably the most important question for investors. Ultimately, the investments need to generate sufficient returns to justify the increasing capital intensity and rising debt levels.
For bond investors, the key point is that the leading hyperscalers already have profitable business models. Existing cash flows from cloud, software and advertising businesses provide a substantial buffer for servicing debt.
Nevertheless, investors should not underestimate the risks. If future earnings from AI investments were to fall significantly short of expectations, this could lead to higher financing requirements and rising leverage over the long term. For us, the ability to monetise these investments therefore remains the key valuation factor.
For bond investors, the key point is that the leading hyperscalers already have profitable business models. Existing cash flows from cloud, software and advertising businesses provide a substantial buffer for servicing debt.
Nevertheless, investors should not underestimate the risks. If future earnings from AI investments were to fall significantly short of expectations, this could lead to higher financing requirements and rising leverage over the long term. For us, the ability to monetise these investments therefore remains the key valuation factor.
What indicators show whether these investments are actually being monetised successfully?
Moritz Krohn: We focus primarily on two metrics: growth in cloud revenues and the development of operating margins.
If rising infrastructure investment is accompanied by growing revenues and stable or even expanding margins, this is a strong indication of successful monetisation.
The latest quarterly results from several hyperscalers show precisely this pattern. This suggests that the strong demand for AI infrastructure is currently not only generating growth, but is also increasingly being translated into economic value.
If rising infrastructure investment is accompanied by growing revenues and stable or even expanding margins, this is a strong indication of successful monetisation.
The latest quarterly results from several hyperscalers show precisely this pattern. This suggests that the strong demand for AI infrastructure is currently not only generating growth, but is also increasingly being translated into economic value.
Will investors need to differentiate more clearly between individual hyperscalers going forward?
Moritz Krohn: Absolutely. From a credit perspective, this is by no means a homogeneous group.
Microsoft, Amazon, Alphabet and Meta currently have very strong balance sheets, high levels of liquidity and substantial financial flexibility. Even if investment levels continue to rise, we see significant buffers within their current investment-grade ratings.
Nevertheless, issuer selection is becoming increasingly important for investors.
Microsoft, Amazon, Alphabet and Meta currently have very strong balance sheets, high levels of liquidity and substantial financial flexibility. Even if investment levels continue to rise, we see significant buffers within their current investment-grade ratings.
Nevertheless, issuer selection is becoming increasingly important for investors.
These companies have issued substantial volumes of bonds over recent quarters. Is there a risk of overheating?
Moritz Krohn: At present, we see no signs of this. Demand from institutional investors for high-quality technology bonds remains strong.
Large issues in recent months were, in some cases, significantly oversubscribed. At the same time, the companies have access to additional sources of financing, such as operating cash flows, infrastructure-related investment vehicles or equity measures.
For the remainder of the year, we expect issuance volumes to decline. Some companies have already largely covered their financing requirements for 2026 and, as part of their financing strategy, have communicated that they do not intend to take on any further debt. Among the hyperscalers that have so far been more restrained — Microsoft and Meta — we believe further bond issuance is more likely. For the other companies, opportunistic financing or pre-financing for 2027 remains possible.
Large issues in recent months were, in some cases, significantly oversubscribed. At the same time, the companies have access to additional sources of financing, such as operating cash flows, infrastructure-related investment vehicles or equity measures.
For the remainder of the year, we expect issuance volumes to decline. Some companies have already largely covered their financing requirements for 2026 and, as part of their financing strategy, have communicated that they do not intend to take on any further debt. Among the hyperscalers that have so far been more restrained — Microsoft and Meta — we believe further bond issuance is more likely. For the other companies, opportunistic financing or pre-financing for 2027 remains possible.
Another issue is so-called off-balance-sheet obligations. Are the risks in this area being underestimated?
Moritz Krohn: The analysis of long-term obligations is becoming more complex and needs to be carried out even more meticulously. Hyperscalers mainly have two types of so-called off-balance-sheet obligations: first, long-term lease agreements that have already been entered into for data centres that will be leased once completed; and second, purchase commitments for hardware components to be delivered in the future.
In our view, however, these items do not currently represent an immediate credit risk. Existing lease obligations become balance-sheet liabilities once the data centre can be used and the company starts generating revenues. As a result, the obligation would only pose a credit risk if the data centre were not sufficiently utilised in the future.
Purchase commitments provide an indication of the capital expenditure that hyperscalers are planning over the coming years. They are reflected in current expectations for capital expenditure.
Both factors nevertheless significantly increase the complexity of the analysis and require a detailed assessment of future cash flows, investment plans and financing requirements. Investors should therefore look beyond the balance sheet and also take account of future obligations and their potential impact on the capital structure.
In our view, however, these items do not currently represent an immediate credit risk. Existing lease obligations become balance-sheet liabilities once the data centre can be used and the company starts generating revenues. As a result, the obligation would only pose a credit risk if the data centre were not sufficiently utilised in the future.
Purchase commitments provide an indication of the capital expenditure that hyperscalers are planning over the coming years. They are reflected in current expectations for capital expenditure.
Both factors nevertheless significantly increase the complexity of the analysis and require a detailed assessment of future cash flows, investment plans and financing requirements. Investors should therefore look beyond the balance sheet and also take account of future obligations and their potential impact on the capital structure.
Is the AI boom changing the structure of credit markets in a lasting way?
Moritz Krohn: We believe so. The technology sector is likely to become increasingly important, particularly in the US dollar investment-grade market.
Our analysis indicates that technology could become one of the dominant sectors in the US credit market by the end of the decade. At the same time, we expect to see increasing concentration among a small number of very large issuers.
In the euro investment-grade market, we view developments as significantly less pronounced. This market is likely to remain heavily shaped by financial institutions, particularly European banks. Even in scenarios involving a significant increase in issuance volumes from the technology sector, we do not expect any fundamental change in the market structure at sector level. However, here too, we are seeing individual hyperscalers gain importance as issuers.
Our analysis indicates that technology could become one of the dominant sectors in the US credit market by the end of the decade. At the same time, we expect to see increasing concentration among a small number of very large issuers.
In the euro investment-grade market, we view developments as significantly less pronounced. This market is likely to remain heavily shaped by financial institutions, particularly European banks. Even in scenarios involving a significant increase in issuance volumes from the technology sector, we do not expect any fundamental change in the market structure at sector level. However, here too, we are seeing individual hyperscalers gain importance as issuers.
What does this mean in practical terms for institutional investors?
Moritz Krohn: Hyperscalers are set to become an even more important part of the global credit universe. Given the high coupons being paid, this also opens up attractive investment opportunities, while at the same time increasing the demands placed on fundamental analysis.
Investors should not focus solely on current earnings power. They should also pay close attention to capital allocation, the development of financing requirements and the ability to monetise AI investments.
The more important these issuers become for credit markets, the more important careful security selection will be.
Investors should not focus solely on current earnings power. They should also pay close attention to capital allocation, the development of financing requirements and the ability to monetise AI investments.
The more important these issuers become for credit markets, the more important careful security selection will be.
Finally, what is the key message for institutional investors?
Moritz Krohn: The key question is not how much hyperscalers invest, but how successfully they can translate these investments into sustainable cash flows.
As long as demand for AI infrastructure remains high and the companies preserve their strong earnings power, we continue to view the major hyperscalers as attractive issuers. At the same time, differentiation between individual companies is likely to increase.
For credit investors, the crucial question is not how much hyperscalers spend, but how effectively they convert AI investment into sustainable cash flows and long-term credit strength.
As long as demand for AI infrastructure remains high and the companies preserve their strong earnings power, we continue to view the major hyperscalers as attractive issuers. At the same time, differentiation between individual companies is likely to increase.
For credit investors, the crucial question is not how much hyperscalers spend, but how effectively they convert AI investment into sustainable cash flows and long-term credit strength.
References to individual securities and companies are for illustrative purposes only and do not constitute a recommendation to buy or sell the securities mentioned. The companies named do not necessarily form part of Union Investment portfolios. Assessments may change, and the company may already have responded to developments.
Source: Union Investment, All information, explanations and representations are as at 30 September 2026, unless otherwise stated