The “second AI winter” refers to the collapse of the commercial AI industry in the late 1980s, after the expert-systems boom of the early part of the decade. This entry dates it to around 1987, the year the market for specialized AI hardware fell apart. As a marker of the field’s own response during this period, this entry cites Raj Reddy’s 1988 AAAI Presidential Address, “Foundations and Grand Challenges of Artificial Intelligence,” delivered as the head of the main AI professional society at the heart of the downturn; the AAAI confirms Reddy as the 1988 president.
The boom that preceded the bust was real. Companies had poured money into expert systems and into special-purpose “Lisp machines,” computers optimized to run the Lisp programming language favored by AI researchers. The trouble was twofold. First, ordinary workstations and personal computers from companies like Apple, IBM, and Sun grew powerful enough to run the same software more cheaply, undercutting the specialized hardware vendors. Second, many expert systems proved expensive to build and maintain and brittle in use, failing to deliver the broad payoffs that had been promised.
The result was a sharp contraction: AI hardware vendors lost their market, many AI startups folded or were absorbed, and funding and corporate interest cooled for years. The episode echoed the first AI winter of the 1970s that followed the Lighthill Report, reinforcing a recurring pattern in which inflated expectations are followed by disappointment and retrenchment.
Primary sourcing note: contemporaneous first-party AAAI documents such as the 1988 presidential address establish the state of the field at the time, but the specific figures often quoted for the collapse (numbers of failed companies and dollars of lost market value) come largely from retrospective secondary accounts rather than original filings, so those precise numbers are deliberately not asserted here.
Why business readers should care: the second AI winter is the clearest historical warning about AI hype cycles. A genuine technology delivered some value, got oversold, and then suffered a backlash that froze investment for years, a pattern worth keeping in mind whenever expectations for a new wave of AI run far ahead of demonstrated results.