AI Has Not Yet Become a Hiring or Productivity Shock

Martha Gimbel read the May jobs report as strong enough to be almost uneventful. Her reaction, she said, was the economist’s version of opening the report, seeing that it was “really great,” finding “not a ton going on,” and moving on.

The labor-market picture was steady rather than fragile. Nonfarm payrolls increased by 172,000 in May, after an April revision to 179,000. Unemployment held at 4.3%, and average hourly earnings growth slowed to 3.4% year over year from 3.6%.

Gimbel’s interpretation was that the report “looks like what it is”: job growth remained strong, while wage growth showed some lag, particularly relative to inflation. For Federal Reserve policy, that combination gives officials room to continue acting on inflation through interest rates.

Mary Daly had framed the same posture as deliberate optionality, saying policy was “in a good place,” the Fed was prepared to respond “either way,” and giving more forward guidance could become “misguiding” if it narrowed what officials were willing to see in incoming data. Ed Ludlow characterized that as fairly standard interim Fed language, but used it to press the question of whether artificial intelligence was already changing the economic data the Fed would need to interpret.

Ed Ludlow put the AI question in two forms: whether it was showing up in productivity data, and whether companies were holding off on hiring because they did not yet know what AI would or would not do for them. Martha Gimbel answered directly: she was not seeing evidence of major AI impacts in the economic data.

That did not mean AI was economically irrelevant. Gimbel separated investment effects from labor-market and productivity effects. On the investment side, she said, AI “is making a difference.” But it did not appear to be restraining hiring, and it did not “really seem to be showing up in the productivity data.”

The investment contrast was visible in Bloomberg’s on-screen table, which paired large announced job cuts with large first-quarter capital expenditures at several major technology companies.

Gimbel’s broader caution was about timing. She called herself “a broken record” on the point, but said it remains “really, really early” for the technology. In her view, expecting AI to have already produced clear labor-market effects risks applying an impossible standard.

Her point was about what is visible in broad hiring and productivity data, not about whether individual companies are cutting jobs or increasing AI spending. The evidence she emphasized separated those layers: AI is already affecting investment, but she did not see it showing up as a major macroeconomic labor-market or productivity force.

The unresolved inflation question is whether the AI infrastructure buildout pushes costs up now, lowers some costs later, or does both on different timelines. Ludlow framed the capital-expenditure numbers as “ginormous” and pointed to a possible disinflationary argument from the utility side: if hyperscalers absorb the burden of large energy and infrastructure projects, energy prices could eventually come down.

Gimbel put limits on the AI explanation before accepting part of it. On energy inflation, she said, the most important factor “right now” was developments in the Middle East, which would “trump anything else.” Within that context, however, she said the inflation data did show “some upward impact” from activity in the AI sector.

Her view was conditional on financing and utility arrangements. If the way utility-side projects are financed changes, she said, AI infrastructure spending “might start working in the other direction.” But she did not think that shift was happening yet.

The harder issue is the feedback loop between rising input costs and rising capex plans. Ludlow pointed to Meta’s earlier capex explanation: one reason for raising capex was not simply a need to spend more, but that the cost of what the company was trying to do had risen because of the broader environment. The question was how circular that process becomes: higher costs lead to higher capex plans, which may feed further into inflation.

Gimbel’s answer was that the AI buildout has the basic ingredients for upward price pressure. The sector is attracting intense investment, and it is expensive to build in. “These models are not free,” she said. “The data centers are not free.” When capital piles into an expensive sector, prices rise; if companies continue investing, they then have to pay those higher prices.

The resulting picture is uneven rather than binary. AI is not, in Gimbel’s reading, a major visible drag on hiring or a clear boost in productivity data. It is visible in investment. It may already be adding some upward pressure to prices. But for inflation, she placed AI beneath larger energy forces for now, and treated any eventual disinflationary effect from infrastructure financing as a possibility rather than a present reality.