The 1990s expert-systems winter
Specialized AI hardware companies such as Symbolics and Lisp Machines Inc., along with expert-systems software firms, attracted heavy investment in the mid-1980s on promises of automating professional work. By the early 1990s the systems were proving brittle in the field, the specialized hardware was undercut by general-purpose workstations, and the entire category collapsed.
Multiple AI hardware companies failed or were acquired at a fraction of their peak valuations. Corporate AI budgets were cut sharply and the term 'AI' became commercially toxic for a decade.
The collapse delayed productive commercial AI by roughly fifteen years and reshaped how investors evaluated deep-tech revenue claims, with lasting skepticism toward research-stage technologies sold as enterprise products.
Today's quantum sector resembles late-1980s AI in two specific ways: specialized, expensive hardware sold against ambitious general-purpose claims, and revenue heavily concentrated in pilot projects rather than recurring production workloads. The relevant question is whether Willow-class error-correction milestones break that pattern in time.