Artificial intelligence can promise cheaper production while making parts of the economy more expensive in the meantime. Federal Reserve Governor Lisa D. Cook put that timing problem at the center of a September 28 speech at Oakland Tech Week, offering a framework for interpreting the investment boom without treating every increase in computing capacity as immediate inflation relief.
In her published remarks, Cook said demand for construction labor and energy associated with data centers can spread price pressures beyond technology. She expects productivity gains to provide modest disinflation over the next few years, but not soon enough to offset broader pressure later this year. She stressed uncertainty about timing and scale. These are one policymaker’s views, not a new interest-rate decision or a guarantee about future inflation.
For businesses and investors, the practical implication is to distinguish the cost of building the technology from the value produced when other companies put it to work.
Spending is visible before the return
A hypothetical company can pay for software, staff training and integration months before it delivers a service more efficiently. During that interval, its spending rises even if the eventual project succeeds. The eventual return must be measured against those costs and against the work needed to keep the system reliable.
That makes announcements of investment and reports of productivity different forms of evidence. An investment plan describes a commitment or intention. A productivity measure asks what was produced with the resources used. Treating the first as proof of the second skips the operational work between them.
The same distinction matters when reading company results. A higher technology budget might support future capacity, improve an existing service or simply replace equipment that is wearing out. The headline amount does not identify which explanation applies.
Follow the benefit through the business
An editor’s test for an AI claim can begin with a concrete task: what became faster, more accurate or less costly? Then ask whether the gain survived review, whether workers could use the saved time productively and whether customers experienced any improvement. A successful demonstration answers fewer questions than a repeatable process used across a company.
Pricing adds another step. A company may use an efficiency gain to lower prices, improve service, fund expansion or strengthen its margin. Those choices have different implications for customers and competitors. A broad claim that AI will reduce costs leaves the destination of the benefit unresolved.
Read the evidence on its own timetable
For a business planning next year’s budget, a useful approach is to separate costs already contracted from savings that still depend on a successful rollout. Record the assumptions behind the savings and identify the evidence that would justify revising them. That avoids turning an optimistic forecast into an unexplained budget requirement.
Cook’s remarks make the sequencing worth watching. The questions for subsequent data are how adoption changes actual output, how widely the gains spread and what businesses do with them. A single technology announcement cannot settle those economy-wide questions, however impressive the underlying system may be.
For the company-level side of adoption, see our coverage of Microsoft’s Copilot push and the questions around usage costs.
Illustrative server equipment; not a specific company investment discussed by Governor Cook. Photo: panumas nikhomkhai / Pexels.