Janus Henderson Credit Team: AI Infrastructure Debt Is Eroding the Diversification Value of Bond Indices
Bradford Smith, John Lloyd, and James Donahue of Janus Henderson note that large-scale investment-grade bonds issued by hyperscale cloud providers for AI infrastructure are automatically flowing into bond indices. This exposes passive fixed-income investors to the same thematic concentration risk in their bond portfolios that they already bear in equities, thereby weakening the traditional diversification benefit between stocks and bonds.
Bond indices have a structural characteristic: they are market-cap weighted, meaning companies with higher debt loads carry greater weight in the index. In an article published on November 4, 2026, the credit team at Janus Henderson points out that this mechanism is producing an unintended consequence alongside the wave of AI capital expenditures.
Capital expenditures by hyperscale cloud providers (Amazon, Google, Microsoft, Meta, Oracle) are projected to exceed $1.1 trillion in 2027, with these expenses primarily financed through the investment-grade bond market. Newly issued bonds are automatically incorporated into indices. The article presents data from two different perspectives: while the pure technology sector currently accounts for approximately 10% of the classification weight in investment-grade indices, including hyperscalers that fall outside the tech classification—such as Amazon (consumer discretionary), Alphabet, and Meta (communications)—the combined weight is already in the mid-teens percentage range. Janus Henderson projects this combined proportion could rise above 20% by 2028. The issue is that when passive bond investors buy into the index, they are effectively accumulating credit exposure to the same group of companies according to set rules—companies that are precisely the source of AI concentration in equity markets.
The article quantifies this change in risk using two dimensions. First is Duration Times Spread (DTS), a metric measuring credit volatility: the DTS of hyperscale cloud providers has become comparable to the weighting of the six largest U.S. banks within investment-grade corporate bonds, meaning their contribution to the overall credit risk of the index rivals that of the financial sector. Second is the spread itself: according to Barclays data, credit spreads for the AI ecosystem were about 10 basis points tighter than the investment-grade average a year ago, but have now widened by approximately 35 basis points, with hyperscalers being the fastest deteriorating group. Widening spreads indicate that the market has begun pricing in this concentration, but index investors cannot reduce their holdings as a result—the rules require them to hold.