The expansion of artificial intelligence has triggered a surge of investment and rising equity values at technology firms. The largest firms, known as ‘hyperscalers’, issued over $100 billion of bonds in 2025, around five times the 2023-2024 annual average. This column compares bond yields for hyperscalers and other firms, before and after ChatGPT’s release in 2022. Hyperscalers’ yields fell by about 29 basis points relative to other firms, and by 42 basis points relative to software firms, whose business models artificial intelligence may disrupt. As hyperscalers’ debt grows, it remains critical to monitor whether these pricing effects continue.
The rapid expansion of artificial intelligence (AI) has triggered an unprecedented wave of investment in the digital infrastructure required to train and deploy large AI models, affecting aggregate investment and production dynamics (Carpinelli et al. 2026, Korinek and McKelvey 2026) and generating substantial implications for electricity demand and energy prices (Bonfiglioli et al. 2025, Ferriani and Gazzani 2026). At the same time, the growing size and importance of large technology firms in financial markets have stimulated an intense debate on whether current valuations are justified by fundamentals or reflect speculative dynamics (Albori et al. 2025, Caballero 2026).
More recently, the footprint of AI firms has expanded beyond equity markets to corporate debt markets. The planned capital expenditure on data centres by the so-called hyperscalers – Alphabet (Google), Amazon, Meta, Microsoft, and Oracle – has risen sharply (Figure 1), straining internal funding, and significantly turning them towards the bond market.
Figure 1 Hyperscalers’ capital expenditure
During 2025, the five hyperscalers issued approximately $108 billion of bonds, compared with an annual average close to $20 billion during 2023 and 2024 (Figure 2). Issuance activity accelerated further during the first months of 2026, with more than $56 billion raised in only a few weeks.
Figure 2 Issuance of US technology firms
Despite the growing literature studying the macroeconomic implications of AI and the effects of AI adoption on firm behaviour (Acemoglu et al. 2023, Fornaro and Wolf 2026, Borri et al. 2026), little is known about how the AI boom is reflected in corporate bond markets. In a recent paper (Albori and Zaghini 2026) we fill this gap by studying whether investors differentiate across firms according to their exposure to the opportunities and risks created by generative AI, using a sample of bond placements by non-financial corporations in the US between January 2020 and February 2026.
To identify the impact of the AI boom on debt financing conditions, we exploit the public release of ChatGPT on 30 November 2022 as a plausibly exogenous turning point that altered investors’ expectations regarding the future distribution of profits within the technology sector. Although artificial intelligence had been developing for decades, the launch of ChatGPT represented a visible and widely recognised turning point in market perceptions regarding the commercial viability, scalability, and profitability of large language models. Importantly, the shock affected market beliefs rather than firms’ contemporaneous fundamentals. The accounting characteristics of technology firms did not suddenly change; what changed was investors’ assessment of future growth opportunities and competitive positioning.
We focus on two groups of firms that occupy opposite positions within the AI value chain. The first group comprises the five large hyperscalers; these firms are central to the provision of cloud computing services, have committed unprecedented amounts of capital to AI infrastructure, and stand to benefit directly from increasing demand for computational capacity. The second group consists of software firms, whose business models may be disrupted by generative AI technologies.
We compare how bond yields changed for hyperscalers versus other firms, after ChatGPT’s release, in a difference-in-differences setting. In our regressions we control for both bond and firm characteristics, as well as for unobservable time-varying factors, to account for macroeconomic trends and the aggregate interest rate environment, like the Federal Reserve tightening period that occurred shortly after our key event.
Hyperscalers have experienced a marked improvement in bond financing conditions during the AI boom, for a total yield discount of approximately 29 basis points, relative to other non-financial corporations. One possible interpretation is that investors viewed AI-related capital expenditure not as a source of additional financial risk but rather as an investment opportunity capable of generating substantial future cash flows. Under this interpretation, expectations regarding future profitability dominated concerns about rising leverage and borrowing needs. We label this yield spread the ‘hyperium’, borrowing from the green-finance literature where the term ‘greenium’ denotes the pricing advantage enjoyed by green securities.
Most notably, our estimate for the discount increases to 42 basis points when we restrict the comparison to software firms, providing further insight into the distributional consequences of the AI shock within the technology sector. Investors appear to have distinguished sharply between firms supplying the critical infrastructure required for AI deployment and firms whose business models may themselves be disrupted by AI-enabled competition. This interpretation is broadly consistent with the emergence of what market participants have increasingly labelled SaaSpocalypse: the concern that generative AI could weaken some of the fundamental sources of competitive advantage traditionally enjoyed by software companies.
Our findings suggest that credit markets have rapidly incorporated expectations regarding the distribution of gains and losses associated with generative AI. Investors appear willing to provide cheaper financing to firms positioned to benefit from the AI infrastructure buildout, while demanding higher compensation from firms facing uncertainty concerning the future viability of existing business models. By lowering financing costs for AI infrastructure providers and at the same time increasing the relative cost of capital for firms more vulnerable to the AI boom, bond markets may reinforce the concentration of investment and market power within a small number of dominant technology companies, while creating refinancing challenges for firms perceived to be on the losing side of the AI transition.
Another key risk arises from the growing stock of debt accumulated by AI-related firms; if expectations regarding AI adoption, profitability, or technological leadership were to change, this could become subject to sharp repricing.
It is thus critical to actively monitor whether these pricing effects persist as AI adoption becomes more widespread and hyperscalers issue more debt, and whether the growing financing needs of AI-related firms generate spillovers to other sectors.
Source : VOXeu
Last year, we surveyed firms in emerging markets backed by private equity (PE) and venture…
For commodity-exporting emerging market and developing economies (EMDEs), recent years have delivered something more unsettling…
Several countries are working to interlink their domestic fast payment systems in order to reduce…
Public services increasingly use predictive algorithms to direct scarce support towards those most at risk.…
Green innovation is central for the transition to a low-carbon economy, but there are concerns…
Explaining comparative economic development requires identifying the causes of institutional change. This column uses AI…