“Digital lettuce” is David McWilliams’s metaphor for AI hardware investment that may lose economic value quickly. The economist and author argues that rapid technological change can make a newly purchased GPU outdated far sooner than investors expect. That is a warning about the useful life of fast-moving hardware—not proof that every GPU, data center or AI investment expires on a fixed schedule.
The distinction matters because data-center buildings and supporting infrastructure can remain useful while the chips inside them are replaced. Ed Yardeni, in a Nov. 20, 2025 Fortune interview, offered that broader-asset counterpoint.
As an Amazon Associate I earn from qualifying purchases.
What McWilliams means by “digital lettuce”
McWilliams said, “You’re investing in something that is a perishable good.” He was describing AI hardware, especially the GPUs used to run and train artificial-intelligence systems, as an asset whose usefulness and market value can deteriorate when a newer generation arrives.
His GPU prediction
In the same discussion, McWilliams said: “Technological change suggests that if you buy a GPU today, the chip is going to be outdated next year.” Futurism’s Nov. 21, 2025 report also recounts that warning.
That sentence is McWilliams’s prediction, not an engineering measurement of every GPU’s service life. A chip can remain operational after a newer model launches; the question is whether its performance, efficiency and cost still make sense for a particular workload.
Is the metaphor about all data-center investment?
Mostly, the dispute concerns the hardware-generation cycle rather than a claim that every data-center facility is about to fail. The arguments in the reporting separate two different assets:
| Investment question | McWilliams’s warning | Yardeni’s counterpoint |
|---|---|---|
| GPU and accelerator hardware | Rapid technological change could make a chip bought today outdated as soon as the following year. | Not the focus of his quoted counterargument; his point is that the wider data-center asset can outlast a particular technology cycle. |
| Data-center facilities and supporting systems | A fast hardware cycle can make the economics of an AI build-out look perishable. | Facilities existed before the generative-AI boom, and some continued operating with their original chips. |
| Accounting useful life | Raises a question about whether assumed depreciation periods match technological change. | Does not establish a universal accounting schedule or replacement rule. |
On that reading, “digital lettuce” applies most directly to rapidly changing GPU capacity and the returns expected from it. It should not be treated as a blanket statement that the land, buildings, power equipment, cooling systems or networking in every data center become worthless when a new GPU generation appears.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Yardeni’s counterpoint: data centers predate generative AI
Yardeni’s response, quoted by Fortune, is: “Data centers existed before AI caught on in late 2022, when ChatGPT was first introduced.” His point is historical and asset-specific: the facilities serving cloud computing and other digital services were already operating before the current generative-AI investment surge.
Fortune also attributes this figure to Yardeni: “During 2021, there were as many as 4,000 of them in the U.S.” That is Yardeni’s reported estimate, not an independently verified official census in the available coverage.
Yardeni further argued that some facilities were still operating with their original chips. That does not disprove the possibility of rapid GPU obsolescence; it shows why a data center should not be treated as identical to the accelerator hardware installed inside it.
Three different meanings of “useful life”
Technical life
A GPU’s technical life is the period during which it functions. McWilliams’s warning is not that a chip physically stops working after one year. It is that a newer chip or a changed workload could make the older one less competitive or less profitable.
Free tools Windows power users keep installed
One-click scans. No signup required.
Economic life
Economic life is the period during which operating the hardware produces enough value to justify its power, space, maintenance and financing costs. This is the sense in which a working GPU could still be “perishable”: its resale value or expected return may fall before it fails.
Best Value
- Family farms not data design for people against AI server farms, data center expansion, rural land buyouts, corporate agriculture, and industrial tech development replacing farmland and open space. Rural conservation and anti data center message.
- AI protest design for farmers, land conservation supporters, anti AI activists, sustainability groups, environmental advocates, rural communities, and people opposing server farm construction, power grid strain, and farmland destruction.
- Lightweight, Classic fit, Double-needle sleeve and bottom hem
Accounting life
Companies choose depreciation schedules for financial reporting and tax purposes. The available reporting does not validate any company’s schedule, show that all operators use the same assumptions, or quantify how a shorter or longer schedule would affect earnings. Accounting treatment therefore cannot be inferred from McWilliams’s metaphor alone.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What “already starting to wilt” does—and does not—claim
The phrase describes a risk that expectations and returns may weaken as AI hardware changes, not evidence that the entire data-center sector is already obsolete. Fortune presents McWilliams as warning of a possible crash in the AI trade; that is a forecast attributed to him, not a confirmed market outcome.
A facility can have several revenue sources and workloads. An operator may refresh only part of its accelerator fleet, move older chips to less demanding tasks, or continue using the building for non-AI computing. Conversely, a site designed around a narrow, rapidly changing workload could face more pressure if demand or pricing falls. The reporting does not provide an inventory or financial study that determines how common either situation is.
What the two reports establish
- David McWilliams, an economist and author, used “digital lettuce” to describe AI hardware investment he considers perishable.
- He predicted that technological change could make a GPU bought today outdated the next year; the statement is his forecast, not a measured lifetime for every GPU.
- Ed Yardeni’s counterargument is that data centers existed before the late-2022 ChatGPT launch and that some still operated with their original chips.
- Fortune attributes the “as many as 4,000” U.S. data centers in 2021 figure to Yardeni.
What remains unestablished
- There is no universal GPU replacement timeline supported by these reports.
- The coverage does not supply an engineering comparison of current GPU models, an audited inventory of U.S. data centers, or a verified count of facilities still using original chips.
- It does not establish company-specific depreciation schedules or calculate the financial effect of changing them.
- It does not show that every data-center investment, rather than particular AI hardware purchases, is destined to become uneconomic.
How to interpret the warning
McWilliams is highlighting duration risk: investors may commit capital on the assumption that AI hardware will earn returns for a long period, while faster innovation shortens the period during which that hardware is competitive. Yardeni is highlighting asset durability: the physical data-center platform can predate and outlast a particular wave of chips.
Both points can be true at once. “Digital lettuce” is a useful caution about fast-changing GPU economics, but it is not a verified expiration date for GPUs or a verdict that all data-center investment is already wilting.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
