The 1990s Chip Glut: A Timeline of the Decade That Reshaped Semiconductors
Archi3 min read
Contents
- The decade the chip industry learned to count
- 1990: The memory market starts to crack
- 1993: The boom returns, and then some
- 1995: Record revenues hide a warning
- 1996: The inventory bubble bursts
- 1997: The Asian financial crisis hits demand
- 1998: The foundry model gains ground
- 1999: Recovery and the 300mm question
- The surprising number
- What the timeline shows
The decade the chip industry learned to count#
In 1996, the global semiconductor industry shipped more transistors than the world had produced in all prior years combined. That milestone arrived just as the business entered its sharpest downturn since the 1980s. The 1990s were a decade of whiplash: record growth, brutal gluts, and a structural shift toward the foundry model that still defines how chips are made.
1990: The memory market starts to crack#
Japanese manufacturers held roughly 70 percent of the DRAM market at the start of the decade. Prices for 1 megabit memory chips began falling in 1990 as new capacity came online. U.S. and Korean producers ramped up output, and the first signs of oversupply appeared. At the time, few analysts expected a prolonged slump.
1993: The boom returns, and then some#
By 1993, the industry was growing at a double-digit clip. PC demand accelerated, and the 486 processor generation gave way to the Pentium. Capital spending rose sharply. Fabrication plants that cost $500 million in the late 1980s were now projected to cost $1 billion by the middle of the decade. Companies borrowed heavily to build them.
1995: Record revenues hide a warning#
Worldwide semiconductor sales reached roughly $144 billion in 1995, a record. DRAM prices were strong for much of the year. But capacity additions in Korea, Taiwan, and Japan were already in the pipeline. The warning signs were visible to anyone tracking equipment orders: too many fabs were being built for the same product categories.
1996: The inventory bubble bursts#
In 1996, DRAM prices collapsed. The 16 megabit chip, which had sold for around $50 at its peak, fell below $10 within months. The industry's inventory of unsold chips swelled. Analysts estimated the global chip inventory reached $15 billion by early 1997. The cause was simple: a wave of new fabs came online just as PC growth slowed.
1997: The Asian financial crisis hits demand#
The crisis that began in Thailand in July 1997 spread across Asia. Currency devaluations made imported components more expensive and slowed consumer demand in key markets. Korean chipmakers, carrying heavy debt from capacity expansions, faced a cash crunch. Several chaebol restructured their semiconductor operations. The industry's total revenue fell in 1997 for the first time since 1985.
1998: The foundry model gains ground#
As integrated manufacturers cut back, pure-play foundries picked up business. TSMC, founded in 1987, had spent a decade building credibility. In 1998, the company reported strong growth as fabless chip designers outsourced production. The idea that a company could design chips without owning a fab moved from curiosity to viable strategy. By the end of the decade, fabless firms accounted for a meaningful share of new chip designs.
1999: Recovery and the 300mm question#
The market rebounded in 1999. PC demand recovered, and the dot-com buildout drove networking chip sales. The next capacity question was wafer size. The industry began planning the shift from 200mm to 300mm wafers, which promised more chips per wafer but required entirely new tooling. The first 300mm pilot lines were announced, with volume production still years away.
The surprising number#
By the time the 1990s ended, the industry had spent an estimated $100 billion on new fabs during the decade. Yet the number that stands out is smaller: 1.5. That was the approximate ratio of capital spending to revenue at the peak of the 1996 building boom. A ratio above 1.0 is unsustainable for any extended period. The industry paid for that overshoot with a two-year downturn that wiped out dozens of companies.
What the timeline shows#
The 1990s established a pattern that persists: capacity decisions made in boom years create gluts in bust years. The foundry model absorbed some of that risk by separating manufacturing from design. But the underlying dynamic, where capital spending outruns demand, has not disappeared. The difference now is that the number of companies able to build leading-edge fabs has shrunk to a handful. When those few firms misjudge demand, the consequences spread through the entire supply chain.
More articles