The Maze: AI will probably change how software, search, advertising, and commerce work. That does not make every AI price rational. The FT/LSEG comparison shared by Ken Kuang shows the part investors forget when a useful technology becomes a market story: Cisco, Amazon, Qualcomm, and Nvidia all survived the dot-com era. Buying near the wrong valuation peak still meant waiting years, and sometimes decades, to get whole.
The survivor list is the trap. Cisco became the world's most valuable company in March 2000, yet the FT/LSEG data marks its return to that valuation neighborhood as `25 years later`. Qualcomm helped define mobile connectivity and still needed `20 years` after its December 1999 peak. Amazon reshaped retail, cloud, and logistics, but the visible recovery note says it took `10 years` after its December 1999 peak. This is the uncomfortable distinction: business truth and investor truth can move on different calendars.
Nvidia is the exception that proves the rule. Nvidia's visible peak is later, in December 2001, and the recovery annotation lands in October 2006. That is faster than Cisco, Amazon, or Qualcomm, but still not instant. The line then surges far beyond the Nasdaq benchmark. The lesson is not that great companies fail. It is that even great companies can spend years digesting a price paid during a mania.
AI has the same split personality. The technology can be real while the capital cycle is messy. AI commerce tools, retail-media automation, agentic shopping, search substitutes, and marketplace seller software may all become useful. But usefulness is not the same as revenue quality, margin durability, switching cost, or payback period. The post's central warning is useful because it separates adoption from valuation, and most AI debate still mashes them together.
Operators should read this as a budgeting lesson, not a market-timing sermon. Ecommerce teams do not need to predict whether AI stocks are too high. They do need to avoid buying every AI promise as if the future has already been de-risked. The smart move is narrower: fund use cases with measurable labor savings, conversion lift, better search, lower support cost, or faster creative testing. Starve the ones that only sell inevitability.
Why it matters: Bubbles do not require bad technology. They require good stories, cheap confidence, and too little math. The dot-com survivors eventually built enormous businesses, but early capital still paid a tuition bill measured in lost decades. AI commerce will create winners. The question for retailers, brands, and marketplaces is whether they are buying capability, or just renting a story at peak price.
Sources: LinkedIn source post | LSEG
