As these chronicles traverse the turn of the century (see related posts below), I'm reminded how the deflation of the dot-com bubble through most of 2000 and 2001 had a significant impact on the semiconductor industry. Specifically for semiconductor test, the large installed base of test equipment largely put in place in the previous few years were then idle. VLSI Research reported that worldwide test utilization reached a local low point of 63% near August 2001 after peaking at 97% just 14 months earlier. For reference, the same source reported another local low of 57% in February 2009 during the recession.

This drop in utilization greatly affected the outsourced semiconductor assembly and test (OSAT) community and overall test provider community. As shown below, the OSAT business model relies on efficiently aggregating customer demand and serving it with an optimally configured and financed capacity mix in order to return profits through economies of scale. Though the OSAT business model is designed to handle the "normal" cycles or swings in the industry (e.g., through flexible "buffer" capacity, equipment leasing, etc.), such a large swing as experienced during the dot-com era caused many to rethink their strategies.

Rather than quickly buying more equipment to grab market share, OSAT leaders were being more cautious and at times pushing back on the demands of the test specifiers. The press reported quotes signaling a new sense of discipline in the market, such as, We remain focused on… improving our asset utilization… [and] on driving… disciplined capital investments" from Tan Lay Koon, CEO of STATS ChipPAC, and We will continue to exercise discipline in our capital investments, with an ongoing focus on improving our productivity" from James Kim, CEO of Amkor. For IDMs, too, this season drove further consolidation and careful analysis of their internal test capacity. The bursting of the dot-com bubble was certainly painful for the semiconductor test industry, but it was beneficial in clearly exposing the challenge unique to the nature of test capacity.

The drive toward more disciplined test capacity investments started after the dot-com bubble.

At this same time, I was still in Austin, but I had just been introduced to a relatively small company (compared to Freescale at the time) out of San Diego called Qualcomm. It was using a Freescale fab in Tempe, Ariz., to manufacture certain RF chips, and some of us from the Teradyne Austin team were called in to try to convince Qualcomm to convert the testing of those chips to the Teradyne Catalyst. This was my first work with a pure-play fabless company, and I was intrigued by what we called the "fabless-subcon" business model. With interest in learning more about this segment of the business, I moved to San Diego in 2002 to work primarily with Qualcomm and other strategic fabless accounts in California.

Over the next several years at Teradyne, working mostly with fabless and OSAT customers, I gained a deeper insight into the extent of the strategic changes resulting from the dot-com bubble. As expected from the aforementioned quotes from OSAT leadership, I saw some large OSAT companies pushing to take more control of the "specification" of their test capacity by promoting select ATE platforms and configurations over others, some even to the point of turning down business with small customers needing unique configurations. Small OSAT companies focused on niche markets requiring a smaller set of configurations in attempts to maintain profitability, knowingly giving up the potential for large-scale growth.

Test equipment suppliers helped the OSAT companies and captive test providers by reducing the number of platforms on the market, but they have more recently added another layer of configuration management complexity with flexible feature licensing schemes. The bottom line is that tightly managing and planning test capacity at high utilization levels remains a difficult and complex task. Tools in the market seem to surround this problem, but they don't quite attack it directly. As I bring these chronicles to the present day next time, I will say more about what is needed for the future of test capacity management.

This article was originally posted at