At IEDM in December, Gary Dagastine is one of the people responsible for press relations for the conference. My piece about Chips and Technologies, the First Fabless Company reminded him that back in that time, he did some work with Jim Dykes on a presentation to Montgomery Securities. Montgomery Securities was one of the three independent San Francisco-based investment banks, along with Hambrecht and Quist, and Roberson Stevens. All three were absorbed into bigger banks in the mid- to late 1990s, ending an era. It turns out that the presentation was exactly 30 years ago today, January 28, 1988. The name Jim Dykes probably doesn't ring many bells, but the company where he was CEO will. For the first year or so of its existence, Jim was the CEO of TSMC. Morris Chang, who would take over as CEO, was chairman of the board (and so Jim's boss). It is an interesting speech to read, since TSMC had been founded just a few months earlier, and so all of the huge success that we know about lay in the future. It is easy to forget that Taiwan as a whole was a country in transition, indeed his friends thought it odd that he had moved there. As Jim said early in his speech: TAIWAN TODAY IS EXPERIENCING A RARE PERIOD OF ENLIGHTENMENT. MARTIAL LAW HAS BEEN LIFTED, NEWSPAPER CENSORSHIP AND CONTROL WAS REVISED ON JANUARY 1, BOATING AND COASTAL FISHING ARE BEING FACILITATED AS RECREATIONAL AND TOURIST ATTRACTIONS, IMPORT TARIFFS ARE BEING SHARPLY REDUCED, AND THE STANDARD OF LIVING IS RISING AT A RECORD RATE. As you can see from the front page of Jim's speech, he titled it The Four Little Dragons of the Orient and an Emerging Role for Semiconductor Companies . The four little dragons were Korea, Singapore, Hong Kong (still a colony of Britain), and Taiwan. The big dragon was Japan. Already you can see just what a different era it was. Japan was scary, China didn't even merit a mention! Jim went into some of TSMC's finances. It has an initial authorized capitalization of $193M and their new fab (fab II) was set for groundbreaking just a couple of months later. Cost of that fab? $220M, which is approximately 1/40th of the cost of a modern fab. I think that is about the cost of two EUV steppers. At that point, the ownership was 48% government of Taiwan, Philips (the Dutch company, at the time with a semiconductor division called Philips Semiconductors, that would become NXP) owned 28%, and the rest some private investors. The vision of what TSMC would become was already locked down at that early point: I MAINTAIN WE ARE NO LESS THAN A PRECURSOR OF AN ENTIRELY NEW WAY OF DOING BUSINESS IN SEMICONDUCTORS. WE ARE A VALUE-ADDED MANUFACTURER WITH A UNIQUE CHARTER...WE CAN HAVE NO DESIGNS OR PRODUCTS OF OUR OWN. T-S-M-C WAS ESTABLISHED TO BRIDGE THE GAP...BETWEEN WHAT OUR CUSTOMERS CAN DESIGN...AND WHAT THEY CAN MARKET. WE CONSIDER OURSELVES TO BE A STRATEGIC MANUFACTURING RESOURCE...NOT AN OPPORTUNISTIC ONE. WE EXIST BECAUSE TODAY'S SEMICONDUCTOR COMPANIES AND USERS NEED A MANUFACTURING PARTNER THEY CAN TRUST...AND OUR APPROACH...WHERE WE AND OUR CUSTOMERS IN EFFECT SPREAD COSTS AMONG MANY USERS...YET ACHIEVE THE ECONOMICS EACH SEEKS...MAKES IT A WIN-WIN FOR EVERYONE. UMC, across the road in Hsinchu, was created a lot earlier than TSMC, and were already a $100M company when Jim gave his presentation. But it would only become a pure-play foundry in the 1990s. So TSMC was the first pure-play foundry, founded in 1987. TSMC had advanced technology, running 2um and 1.5um, with a path to 1.2um and submicron by 1990. That was roughly two process generations behind the most advanced IDMs—although we didn't call them that back then, they were just semiconductor companies. Their total capacity was 10,000 6" wafers per month. That is equivalent to 2,500 12" (300mm) wafers per month...a modern fab has capacity more like 50,000 300mm wafers per month. In total, TSMC manufactured over 10 million wafers in 2016. The big question investors asked was apparently "Is there really a market for all of this capacity? Who needs it?" Of course, having seen how the fabless revolution has played out, this is almost comical, but a lot of people were unconvinced that the model would work. Remember that only a couple of years earlier, Chips and Technologies had difficulty raising money since investors couldn't understand the concept of a fabless semiconductor company—and their first real money came from a real-estate investor who didn't know the difference. TSMC understood though, and one of the five categories of customers for TSMC were "so-called fabless start-up companies." When TSMC polled customers about issues, the most common response was "My foundry should be my partner, not my competitor." Even in 1987, the relationships were starting to mature, with foundries becoming strategic manufacturing partners: COMPANIES AND THEIR FOUNDRIES WILL SHARE CAPITAL PLANS AND PROCESS TECHNOLOGY STREAMS. COMPANIES AND THEIR FOUNDRIES WILL OBJECTIVELY EXCHANGE TECHNICAL KNOWLEDGE...RESULTING IN FASTER AND BETTER ENGINEERING DECISIONS. SOPHISTICATED CUSTOMERS WILL ESTABLISH DEDICATED INTERNAL ORGANIZATIONS TO MANAGE FOUNDRY QUALIFICATION AND PRODUCTION. THESE NEW AND STRONGER RELATIONSHIPS WILL SET A PATTERN FOR THE SUCCESSFUL USE OF FOUNDRIES IN THE FUTURE. TSMC was the first pure-play foundry. But others were watching. UMC later transitioned, but very soon after TSMC got started, Singapore announced the creation of Chartered Semiconductor (now part of GLOBALFOUNDRIES): WE FEEL T-S-M-C'S THEME OF "PARTNER, NOT COMPETITOR" HAS TRIGGERED AN ENORMOUS RESPONSE FROM THE MARKETPLACE. OUR COOPERATIVE APPROACH TO MANUFACTURING IS BEING VALIDATED...NOT ONLY IN TERMS OF OUR SALES OPPORTUNITIES..BUT BECAUSE WE SEE OTHERS COPYING THE IDEA. I AM REFERRING TO THE RECENT ANNOUNCEMENT OF CHARTER [sic] SEMICONDUCTOR...TO BE BASED IN SINGAPORE. Foundry and Fabless Win It is interesting to see how little has changed in 30 years. The basic idea of a foundry with no product lines of its own, and fabless semiconductor companies, and partnering with customers in long-term manufacturing relationships, were all in place from the start. Of course, the numbers for investments, process geometries, wafer capacities are all amusing. Moore's Law is exponential, and Moore's so-called second law, about the cost of fabs doubling every four years, is also exponential. I just looked on TSMC's website and one of the things that remains consistent all the way back to this presentation 30 years ago is: 5. Treating Customers as Partners Since the company was founded, we have treated our customers as partners and have never competed against them. This policy is the key to our current success and will be crucial to our continued growth. At TSMC, customers come first. Their success is our success, and we value their ability to compete as we value our own What was not so obvious back then, 30 years ago today, was just how dominant the fabless/foundry model would become. Foundries and fabless companies feature in the top 10 semiconductor companies today (yes, there is some double counting going on, but that is a second-order error). I think that 30 years ago, a lot of people assumed that when the fabless companies grew up, and had enough demand, they would build their own fabs. But nobody ever did. Foundries' victory was total. And far and away the most successful foundry is TSMC. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.Image may be NSFW.
Clik here to view.![]()
Clik here to view.