|
Major technology companies financing artificial intelligence (AI) infrastructure still have significant room to raise debt despite recent volatility in AI-related bond markets, with J.P. Morgan estimating that hyperscalers could collectively issue around USD 1.7 trillion of additional high-grade debt before reaching levels that could attract greater investor concentration concerns.
The report said recent weakness in AI-related bonds reflects investors demanding higher pricing for new issuances rather than concerns that the market has reached its capacity to absorb additional debt. "The current hyperscaler widening is a byproduct of the HG investor community trying to rationally price in an accelerating pace of issuance from the hyperscalers and not a reflection of insurers already hitting issuer risk limits for these issuers," the report said. According to J.P. Morgan, demand for hyperscaler and associated data centre financing is expected to remain robust, although pricing is likely to stay dynamic as expectations around future funding requirements and monetisation strategies evolve. The report added that investors have become more aggressive in dictating pricing and terms in primary debt markets as issuance has accelerated. The report noted that AI-related debt has come under pressure in recent weeks, with high-grade and high-yield high-performance computing bond indices widening sharply. However, it said the recent repricing is driven more by pricing adjustments than by any shortage of market capacity. It also observed that the broader high-grade market remains far less concentrated than during previous credit cycles, while insurers continue to have room to add exposure to hyperscaler debt. J.P. Morgan further said the AI and data centre financing universe continues to expand rapidly, now comprising 31 issuers with more than USD 576 billion in outstanding bonds and over USD 5 billion in leveraged loans. It expects this investment universe to continue growing as AI infrastructure spending increases. The report also highlighted that securitised data centre products have been relatively resilient compared with corporate AI-related bonds, while adding that the divergence between data centre project bonds and hyperscaler bonds reflects an orderly repricing by investors rather than a deterioration in market fundamentals. (ANI)
|
