AI infrastructure debt pushes tech past banks in major investment-grade bond indexes

Technology accounts for more than 10% of the Bloomberg U.S. Corporate Bond Index, and in several major investment-grade benchmarks it outweighs banks for the first time, according to ETF Trends. The cause is a change in who borrows, because the AI buildout has turned the largest technology companies into some of the corporate market's most prolific issuers of long-dated debt.

The scale is steep. Alphabet, Amazon, Meta, Microsoft and Oracle together sold an average of roughly $35 billion of bonds a year between 2020 and 2024, then $93 billion in 2025, and through the first half of 2026 had already passed $132 billion, six months that outran any full year in the preceding five. Industry forecasts cited by ETF Trends put total AI-adjacent issuance in 2026, a grouping that also takes in chip makers, data centers and power utilities, at $300 billion to $570 billion.

Debt-weighted indexes do not deliberate. A bond benchmark that weights by outstanding debt buys more of an issuer as that issuer borrows more, the fund tracking it follows, and that ordinary arithmetic is why ETF Trends describes investors in conservative corporate bond funds as absorbing concentration in the hyperscalers without having chosen it. The reporting does not itemize per-issuer weights, so the size of any single tech position inside a given fund is not established by the coverage; what is established is the direction and scale of the sector move.

The clearest evidence sits in the funds that own nothing but credit, where ETF Trends puts technology weight in the iShares iBoxx Investment Grade Corporate Bond ETF (LQD), the Vanguard Intermediate-Term Corporate Bond ETF (VCIT) and the iShares 5-10 Year Investment Grade Corporate Bond ETF (IGIB) at 13% or more, up roughly 300 basis points from about 9% in 2024. Oracle, Amazon and Meta now rank among top corporate holdings alongside money-center banks JPMorgan Chase and Goldman Sachs.

The iShares Core U.S. Aggregate Bond ETF (AGG) and the Vanguard Total Bond Market ETF (BND) show the same movement at smaller scale: they remain mostly Treasuries and mortgage-backed securities, with technology at just over 4% of the portfolio, up from 2%, and the corporate slice inside them has traveled the same path, according to the same reporting. A 4% position inside a portfolio anchored by government and agency paper is a different proposition from 13% inside a fund that holds only corporate bonds. For a client whose fixed income allocation is split between a total-market fund and a credit fund, the concentration question lands almost entirely on the credit sleeve, which is usually the smaller of the two.

ETF Trends is careful not to treat the shift as an alarm, citing hyperscaler operating cash flow and credit quality, and its distinction between added exposure and added concentration is fair because for a debt-weighted index the two come from the same arithmetic. What that arithmetic cannot do is ask whether the resulting portfolio still serves the reason a client bought a core bond fund in the first place. An investment-grade allocation typically earns its place by behaving differently from the equity market in a drawdown, yet the debt raised to build data centers, custom silicon and grid connections is a claim on the same capital-spending cycle that drives the index's largest equity positions. Whether the two sleeves now fail together is an empirical question, and the next credit downturn is what answers it.

None of this happened because the funds changed their construction: an index fund delivers the composition of the market it tracks, and the market's composition is set by the borrowers rather than the buyers. The practical consequence is that an investor who wants investment grade credit without the AI capital cycle now has to express that as a choice, because the passive default no longer avoids the question on its own.

PWD has followed the same spending cycle from the equity side, including WGMI's rewrite of its index rules to run from bitcoin miners toward AI power. The credit version is quieter and, for a core bond allocation, harder to see: no mandate was rewritten, no manager made a call, and the exposure arrived through the benchmark.

That range cuts both ways, and the low end deserves as much attention as the high end. If AI-adjacent issuance lands anywhere near $300 billion in 2026, corporate benchmarks add technology weight by arithmetic alone, and the share of the index tied to the buildout keeps climbing without a single investor choosing it. A rating action or a single large issuer stepping back from the market would not undo years of accumulated debt, but a slower 2027 issuance calendar would eventually show up as a flattening sector weight, and a shortfall against the forecast would show up as a falling one. Debt-weighted indexes shed an issuer as its bonds mature and are not replaced, with the same indifference that built the position.

Core bond allocations are sold on the promise of ballast, and that ballast is now partly a credit view on data-center capex, whether or not anyone wrote that down. The practical question is not necessarily whether to sell LQD but which of a client's portfolios is carrying the AI buildout twice, once in the equity sleeve and again in the credit sleeve that was bought to offset it.