Silicon Valley’s Wars for Global Tech Dominance
The earliest organizational wars in chips were Silicon Valley's Upstarts against the East Coast Giants which occurred in the first half of the twentieth century. By the early eighties, the silicon wars between the U.S. and Japan were raging, with Silicon Valley deep in the middle. There are a lot of lessons to be learned in how American companies turned the tables organizationally on Japan. Taiwan’s success in semiconductors also has the underpinnings of the approach with its foundry model.
While the analogy is distasteful, reading Stanley McChrystal’s book Team of Teams, I saw that the East Coast Giants (ECGs) were a lot like the U.S. Military with highly optimized and efficient top-down organizations. The Silicon Valley Upstarts (SVUs) were more like Al Qaeda: relatively inefficient with poor resources and yet deadly flexibility to respond and capture emerging opportunities.
Eventually, these management rebels even dressed differently, exemplars such as Bob Noyce appearing in Fortune magazine with an open collar and the necktie replaced with a gold chain. Their flexibility and speed to develop came from their organizational flatness and egalitarian structures. These were blue ocean strategies, as there were no way higher ups at the ECGs would give up their perks or the perceived profits they robbed from inventors by giving them a small check and plaque for every patent granted. In contrast, the SVUs gave stock options, so those creating the value shared the wealth creation with investors. Investors shared short term gains, but lost far greater long-term ones.
U.S. versus Japan: Later came the silicon wars between the U.S. and Japan. The SVUs had the market share and far better sales ability. But they were no longer upstarts and had grown arthritic with infamous internal turf wars. Fabs in a company competed against each other, while accepting little from R&D. At Intel, they coined the term 'refsnart.' Used to describe the process of manufacturing reverse engineering anything R&D threw over the wall; the word was simply 'transfer' spelled backwards. At AMD, no one from R&D was allowed into a fab without the head of operations' written permission. Essentially, R&D needed a hall pass to get near a fab. At Motorola, new fabs went underutilized as product managers filled R&D with production wafers to avoid the depreciation charges that were added when they used the new fabs.
Japan came to the market organized as an industry by government. Focused on DRAM, heavily structured, well financed, and emboldened by the predictability of Moore’s Law; they moved forward as an unstoppable army. Transfer was never a problem, as they ran multiple R&D teams that would stay with a generation well into manufacturing before rolling back to restart with a new generation. But their very strengths would become weaknesses once the market became unpredictable.1 American companies turned the tables with highly structured tech transfer processes (such as Intel’s famous ‘Copy Exactly’) while pivoting out of memory and into markets such as processors and SoCs, where the need for short decision cycles put Japanese companies at a disadvantage.
American IDMs versus Taiwan: America’s move to heavily structured, top-down control over the fab would leave an opening for Taiwan, which was entering a highly competitive world with neither the marketing or financial resources to be on a level playing field. Their advantage would come from an organizational ‘Team of Teams’ approach where lots were stewarded through the typical fab traffic jams like cars moving through an intersection in Taipei: No one stopped, everyone looked for the open tool, and the simple rule of never crash into someone else. Similar to what McChrystal describes, they had empowered decision making deep into the fab, where all the decision makers had a culture of shared purpose. The advantage also came from the structural big bang of EDA making the fabless/foundry model possible. At the time, this was killing ASICs with standard cell techniques that would give rise to the SoC. Fabs had to be much more flexible and TSMC and UMC adapted best to these changes.
1 I could be biased on this point, as it was a weakness I used back in the eighties to advise clients trying to figure out how to compete against them. I believed that the long decision cycle times, communication barriers to customer needs alignment, and a general inflexibility in strategic pivots would make Japanese companies unable to compete effectively in markets such as processors and what came to be SoCs; where IP development as well as fluency and adaptability to standards made short decision cycles essential. Having CxO execs who could make decisions in front of customers is a key reason why companies like TEL, JSR, and Advantest came out on top.
Originally published September 11, 2015, The Chip Insider
Author: G Dan Hutcheson