AI’s borrowing boom: What it means for bond indices

For investors, IG debt offers an alternative route (via contractual income and a claim on enterprise cash flows) into the AI buildout, versus or alongside equity participation. As of end-September 2026, hyperscalers accounted for 4.7% of the FTSE US BIG Corporate index and 1.2% of the FTSE Euro IG Corporate index by market value weight (up from 2.7% in the USBIG Corp and 0.08% in the EuroBIG Corp indices 2 years earlier). 

The sudden surge in issuance from this segment can broaden and diversify corporate bond sectors while simultaneously concentrating portfolios in a relatively small group of borrowers sharing the same investment cycle. While hyperscalers’ debt still has a relatively small index footprint, it is now significant enough to impact top-line index performance. 

Collectively, the US dollar-denominated bonds of the hyperscalers underperformed the FTSE US Broad IG Corporate index over the 5 years to September 2026, with higher annualised volatility (see Table 1). Excluding hyperscalers from the index would have slightly improved its performance over the period. 

Source: FTSE Russell, unhedged US dollar returns from September 2021 to September 2026.  Past performance is not a guide to future returns.

On average, hyperscalers now have a higher duration, yield and coupon than the FTSE US Broad IG Corporate index. However, while the five hyperscaler issuers have an index weighted average A+ rating, the stand-out issuer is the higher-yielding Oracle, with an average BBB- rating, shorter average life, lower duration and option-adjusted credit spreads (OAS) currently around 100 basis points (bps) higher than its peers (see Table 2). 

While Oracle is a cloud provider, it differs from the rest of the hyperscalers in having entered cloud computing after the initial, web-driven horizontal scaling era. As a result, Oracle has been raising substantial debt since 2016 to “catch up” [note6]. It was downgraded by S&P from A to BBB+ in 2021, to BBB in 2022, and then again to BBB- (the lowest investment grade credit rating) in 2026.

Source: FTSE Russell, data as of 30 September 2026. Past performance is not a guide to future returns. Please see the end for important legal disclosures.

While Microsoft, Alphabet (Google) and Amazon trade with OAS below the FTSE USBIG Corp index average, Oracle's bonds trade at wider OAS than the other four issuers (see Chart 3). Investors are already demanding more compensation to take on Oracle’s debt, but the prospect of further downgrades (resulting in junk status) would likely raise investor concerns further. As of 21 October, the five-year credit default swap (CDS) spreads for Oracle stood at 331 bps,  versus 80 bps for Amazon and 56 bps for Microsoft, representing a five-year default probability of 25%, 7% and 5%, respectively. In particular, Oracle's CDS spread is now well above the typical levels for a BBB-rated issuer (~150-200 bps), suggesting the market may be pricing in downgrade risk already .

However, in the event of an S&P downgrade to below investment grade, Oracle would not necessarily become ineligible for IG indices; this would depend on the rating logic pertaining to the specific index provider. 

In the case of the FTSE USBIG Corporate Index, Oracle would have to lose its investment-grade rating from Moody’s (the agency’s lowest investment grade rating is Baa3, while Oracle’s current rating is Baa2), as well as its BBB- rating from S&P, for its bonds to be removed from the index. This is because, in the case of a split rating between the two credit rating agencies, FTSE Russell takes the higher rating. However, in the case of its removal from the FTSE USBIG Corporate index, the firm’s debt could be eligible for the FTSE US High Yield Market index, which includes bonds from issuers with ratings ranging from BB+/Ba1 down to C/Ca (excluding defaults).

Other index providers may use different rules to assess eligibility for investment grade corporate bond indices, particularly in the case of split ratings, so it’s worth checking the ground rules of the index concerned.

The widening of hyperscalers’ credit spreads could, in theory, be due to perceived “dot-com”-style red flags in the AI industry, such as circular financing arrangements, execution risks and rising borrowing costs. However, the spread widening may also reflect the massive influx of supply to the US IG bond market (which in turn has driven the need by hyperscalers to diversify their bond issuance across currencies). 

As is shown in Chart 4, hyperscalers (as a simple market value-weighted cohort) previously traded at a lower OAS than the broader FTSE US BIG Corporate index, but since November 2025, this has reversed and their spreads now trade above it. By October 2026, hyperscalers’ OAS stood at 119 bps, compared with 86 bps for the broader index—a premium of 33 bps. Since December 2025, the group’s OAS has widened by approximately 31 bps, while the index’s OAS has widened by 1 bp.

The divergence illustrates how headline index spreads can mask substantial changes within issuer groups, reinforcing the importance of looking beyond broad-market averages to understand the exposures embedded in corporate bond benchmarks.

However, the extent of hyperscaler credit spread changes has differed across currencies. In Chart 5, the blue line represents the euro-denominated hyperscaler issues, whose credit spreads have mostly tightened since April 2026 (the first euro-denominated issue by Amazon occurred in March 2026), while the US dollar bonds’ spreads (the red line) show a large widening over the same period. These spread comparisons are term bucket-adjusted, but this is not quite a like-for-like funding cost comparison due to residual differences in duration, issuer weights, bond liquidity and so on.

Is there any danger of AI-related bond issuance tipping the existing status quo of sector allocation within typical IG corporate indices? For example, ‘Banking’ currently makes up 21.5% of the FTSE US IG Corporate Index by market value weight (see Chart 7) and its sector weight has remained mostly stable over the last couple of years. 

In contrast, the total weight of ‘Information & Data Technology’ in the FTSE US BIG Corporate Index has risen from 4.2% to 6.3% in the year from October 2025 to October 2026, making it the third-largest FTSE Russell corporate bond sector (COBS) after ‘Electric Utility’ (which has a 9.1% weight). 

However, the fact that Amazon is considered a retail issuer under both the FTSE COBS and Thomson Reuters Business Sector (TRBC) classification schemes, while the other four hyperscalers are classified as technology issuers (‘Information & Data Technology’ under FTSE COBS and ‘Software & IT Services’ under TRBC Business Sector) is a strong illustration that a simple sector grouping could lead investors to miss parts of the chosen cohort. 

Additionally, while these 5 hyperscalers currently remain the largest source of AI-themed bond issuance, such issuance is likely to show up more broadly over time in a variety of sectors: for example, technology, communication services, utilities, infrastructure, data centres, telecoms and some industrials. New research into identifying those infrastructure suppliers, applications/platforms and adopters will be required to help investors allocate accurately to this fixed income market theme.

If hyperscalers’ share within IG corporate benchmarks continues to grow, their spread moves will contribute more to indices’ headline performance, whether positively or negatively. A clear general implication is the need for clearer exposure measurement and a thoughtful, deliberate choice of benchmark. Indices need to have the supporting data for different types of investors: those requiring broad beta, those concerned about concentration constraints and those seeking deliberate thematic exposure to AI. 

By design, market value-weighted bond indices assign larger weights to issuers with more eligible debt. While this generally provides liquidity, diversification can no longer be an automatic assumption. Previous research from FTSE Russell highlights the growing importance of alternatively weighted bond indices, like the FTSE Debt Capacity World Government Bond Index (DCWGBI), which adjusts market value weights using debt capacity measures.

To construct a fixed income index that allows investors targeted, direct exposure to all AI-themed issuers, new sector classifications that adequately capture the diverse nature of these businesses will be required. More granular, equity-like sector classifications, such as the TRBC sector classifications, could be a more effective tool for thematic investment in fixed income – but alone, they risk missing out key names in the AI ecosystem. 

Either way, the demand for more complex, nuanced fixed income index solutions for passive investment strategies has never been greater.