BIS finds a brief Treasury premium rise around AI financing news

The estimated premium investors demanded for holding ten-year US Treasury risk rose by about 1.5 basis points on major AI financing-news days, then reversed within two trading days, according to a BIS study published on 27 October. The result suggests a brief market association, with no statistically significant change in investment-grade corporate spreads.[1][2]

The comparison adjusts for two-year yield changes and Federal Reserve policy-statement days; the chart uses 90% confidence bands. The bulletin puts a typical daily term-premium move at about 3.5 basis points, providing scale for the 1.5-basis-point news-day association. Anticipated announcements and coinciding developments prevent a causal reading.[1][2]

A modelled component of the borrowing rate

The measure comes from the Adrian–Crump–Moench framework, developed at the New York Fed. It separates a fitted Treasury yield into the expected path of short-term rates and a premium associated with holding interest-rate risk. That distinction matters because a long-term yield can rise for different reasons: investors may expect higher short-term rates, demand more compensation for holding a longer bond, or both. The premium is inferred statistically; it is not a separately quoted market price.[3]

The original methodology uses a sequence of regressions to estimate how yield-curve factors evolve and how their risks are priced. Its daily application allows researchers to examine movements around particular dates. For readers, the consequence is straightforward: an estimated increase in this component is not automatically the same increase in the total Treasury yield, still less an observed change in a particular company’s loan rate.[3]

Two kinds of uncertainty

A Federal Reserve analysis published in 2017 provides relevant methodological context. Comparing term-premium models, it found that including surveys of interest-rate expectations could change their estimated levels and movements. Surveys can improve precision while introducing their own assumptions. The authors specifically cautioned against placing too much weight on short-run premium changes when overall yields move little. That is a warning about the measuring instrument, separate from the difficulty of attributing a market movement to one announcement.[4]

The result is a reason to investigate a possible transmission channel, not to add 1.5 basis points mechanically to a company’s financing forecast. The consequential questions are whether the pattern persists under other term-premium measures and whether sustained AI borrowing changes the lending terms businesses actually receive.

Report an error or suggest a correction