Right here’s the textual content of a be aware to purchasers simply revealed by Jonathan Hill, Barclays’ Head of Inflation Analysis Technique, about why “the inflation threat premium needs to be unfavorable past the close to time period”.
Whereas monetary markets are laser-focused in the mean time on geopolitical dangers and the near-term inflationary penalties of the spike in vitality costs, beneath the floor, a rising consensus is forming that core US inflation dynamics are shifting again towards their pre-COVID configuration. But when we take that baseline critically—and overlay the quickly rising chance of significant labor market disruption from generative and agentic AI—the conclusion could also be uncomfortable for the inflation market. Markets ought to arguably be pricing a unfavorable core inflation threat premium past the close to time period, not a optimistic one, which might be a flip from our considering getting into this 12 months and underscores how shortly dynamics are altering.
Earlier than turning to AI, think about the underlying arithmetic of inflation. Earlier than the COVID-19 pandemic, the US core CPI ecosystem was remarkably secure. Core items inflation oscillated round 0%. Lease inflation ran roughly 3.0–3.5%. Core companies ex-rents sat between the 2 in a means that delivered a weighted common of about 2%. This distribution held for years and offered the anchor for coverage and market pricing.
Now, if we assume tariff pass-through fades and core items return to one thing like 0% in H2 2026, the talk shifts squarely to the opposite two parts. Lease inflation is already extensively anticipated to run under its pre-COVID trajectory. Incoming information from new lease measures, development pipelines and emptiness charges all level in the identical path. If rents are now not delivering 3.0–3.5% and core items inflation returns to flat, then basic math requires core companies ex-rents to run greater than they did pre-COVID simply to maintain the core basket close to 2%.
This a part of the inflation debate has not obtained sufficient consideration, in our view. The bar for supercore inflation is now greater. But, wage monitoring—a major enter into supercore—has already decelerated to its pre-COVID tempo. The Atlanta Fed wage tracker and Barclays’ state-space decomposition present wage momentum sitting at or barely under the place it will must be to generate 2% inflation on this new regime, given softer rents and flat items.
This raises the uncomfortable query: if wage development is already too mushy for the brand new core inflation arithmetic, what occurs if productivity-enhancing or labor-substituting applied sciences speed up?
That brings us to generative and agentic AI.
Over the previous six months, the step-function enchancment in mannequin functionality—and the believable emergence of agentic architectures that may autonomously full multi-step duties—has shifted the chance distribution of labor-market outcomes. The viral weblog publish final week resonated exactly as a result of it articulated a extensively felt, however poorly formalized, instinct, particularly that we could also be approaching the purpose at which AI substitutes not simply duties, however workflows. The response throughout industries speaks for itself. Block’s announcement of serious layoffs—explicitly tied to automation—gave that instinct company type.
Dismissing these developments on the grounds that there isn’t any proof within the labor information misses the underlying dynamics. That argument is analogous to saying hurricane forecasts are incorrect as a result of there isn’t any present flooding. By the point the labor market exhibits stress, the repricing will have already got occurred. Markets don’t look ahead to realized outcomes when the distribution itself has shifted.
One other critique—that the instruments usually are not totally there—additionally feels more and more hole. The metaphor is just not the gradual emergence of a mature expertise; it’s the comedic second in The Workplace when Michael and Dwight comply with the GPS right into a lake. The software program was not good, however the system nonetheless redirected conduct in ways in which had actual penalties. A sober evaluation of the place AI techniques might be within the subsequent a number of years, based mostly on noticed enchancment curves and introduced roadmaps, suggests sharply rising chance of significant disruption, which may disproportionately weigh on wage momentum.
If the market begins to internalize the likelihood that AI may soften labor demand earlier than it meaningfully lifts productiveness, then the danger to inflation is skewed decisively decrease. And the present pricing of inflation compensation doesn’t totally replicate these draw back dangers. Furthermore, at a sure level, the evolving AI shock ought to grow to be a optimistic productiveness shock, which needs to be disinflationary, a minimum of within the medium time period; this level is extra compelling than the extra dire unfavorable mixture demand situation specified by among the extra alarming prognostications.
And but, 2yfwd3y CPI swaps stay above 2.3%, whereas 2yfwd3y actual price swaps solely briefly dipped under 1% and are nonetheless 20bp above their lows from final September. If buyers start assigning higher weight to the AI-disinflation situation—whereby wage development undershoots simply because the supercore hurdle has risen—the response perform throughout charges markets may shift shortly. A downward repricing of inflation threat premia would probably push the entrance finish of the nominal yield curve right into a rally, compress actual yields, re-steepen the curve, and break the present low-volatility equilibrium that has outlined markets in latest months (although that’s already in danger, given the battle in Iran). The near-term dangers are extra two-way, as exemplified by the vitality value shock that drove front-end inflation greater on Monday on a spike in geopolitical tensions within the Center East, to say nothing of large AI capex that may very well be supportive for inflation stress, or broadly rising electrical energy costs. That 1y1y CPI swaps truly fell barely on Monday may point out how a lot of a read-through there could find yourself being from the Iran state of affairs to underlying inflation momentum.
The query is just not whether or not AI has already disrupted the labor market; it’s whether or not the chance of such disruption is rising quick sufficient to affect the distribution of future inflation outcomes. On that dimension, the proof is more and more troublesome to disregard.
Oh, and as a closing level… this piece was largely drafted by AI, executing an intensive immediate laying out the arguments we needed to make and examples we needed to make use of in a single iteration, with solely slight tweaks and edits required afterwards. Whereas analysts will stay the supply of ideation and wish to write down their ideas within the type of a complete immediate, AI is enabling higher and quicker execution.
