(Back to routine writing after my first reflection last week. I intend to do more such reflections in due course. I am excited to see how it comes about)
This post has been coming for some time. Semiconductors are so ubiquitous that the current shortage we are witnessing is creating geopolitical tensions globally.
The United States could not have been more subtle with the timing of the erstwhile Endless Frontier Act, now US Innovation and Competition Act. Should you wonder where the retired moniker came from, it’s inspiration goes all the way back to 1945. Indiana Senator, Todd Young introduced the bill that would help steer nearly $100 billion over 5 years to develop and innovate new technology through business-academia partnerships. Here is Young, in no uncertain terms, talking about the bill.
Americans have always looked towards the frontier and forward to new horizons. This bill, this moment, it’s not only about beating the Chinese Communist Party; the Endless Frontier Act is about using their challenge to become a better version of ourselves through investment in innovation.
Young’s Democratic counterpart from New York in this bipartisan bill, Chuck Schumer expanded its scope to authorise a few key spending initiatives including an outlay of $49.5 billion over five years to establish a “CHIPS for America Fund”.
China setup its second semiconductor focused investment fund in 2019 raising ~ $29 billion, taking the total corpus of both the funds to ~$48 billion. In 2020 alone, cash flowing into Chinese semiconductor firms stood at ~$35 billion through the primary and secondary markets, a 407% increase y-o-y. China, despite being the largest consumer of semiconductors, produces only 5.9% of global output. If there was enough noise made about Chinese reliance on imports, it has reached a fever pitch over the last couple of years when US started holding it as leverage against Chinese-domiciled companies like Huawei.
The European Union has announced its ambition to manufacture up to 20% of the global high-end semiconductor output by the year 2030.
It is clear that powerful nations across the world are getting nervous when it comes to semiconductors. The shortage started with the onset of the pandemic, but tensions have been escalating for at least a decade. And it goes beyond global semiconductor supply chain.
There are only a few companies that can fab a semiconductor at-scale. It is also safe to assume that a similar number can supply the hi-tech equipment and technology to manufacture them (like ASML in the Netherlands). I read somewhere that if one was to strategically go after the corporations that possess the technological nous and take them out, the world would be set back by at least a couple of decades.
It does sound a wee bit unfair to say that developed economies should have exercised better foresight in preempting this ‘capability concentration’, especially given we are talking about the building blocks of the digital economy.
But it makes for an interesting story about how an unassuming man in his 50s turned the semiconductor industry on its head.
Morris Chang was born in the year 1931 into a middle-class family in Ningbo, a port city south of Shanghai. The Chang family moved around a lot during his early years owing to his father’s government job. It was not an easy time to be living in China with visible violence and widespread poverty. Encouraged by his parents to focus on school and academics during this tumultuous period, Morris worked his way to an admit to Harvard. He arrived in the US in the year 1949. He later transferred to MIT and graduated with a Bachelor’s and a Master’s degree in Mechanical Engineering.
He could not clear his PhD qualifying examination subsequently and he took an entry-level position with Sylvania Semiconductors in 1955. Here is a fascinating anecdote from around that time that speaks to Chang’s drive to be excellent at what he did.
His first assignment was to work on germanium transistors, first on improving manufacturing yields and later on developing devices. Chang started working toward yield improvement by looking at bonding, the process by which electrical contacts are attached to the transistors. At the time, a technician on the assembly line connected those electrodes by soldering on a wire, and the heat often damaged the transistor. Chang devised a way to connect the wire using indirect heat instead.
It was a basic mechanical engineering problem; he didn’t have to understand semiconductor theory to figure it out. But he started studying semiconductors anyway, using as his text Electrons and Holes in Semiconductors with Applications to Transistor Electronics (1950) by William Shockley.
Parts of the book baffled him. But he soon discovered that a senior engineer at Sylvania spent hours every evening drinking heavily at the bar in the same hotel where Chang was living, in Ipswich, Massachusetts. So Chang’s typical evening in those days went like this: He’d spend a couple of hours holed up in his room poring over a section of the book and trying to solve the problems on his own. Then he took his questions down to his friend in the bar. “He didn’t solve all my problems, but he solved enough so I could move ahead. He was my main teacher,” Chang says.
He started designing transistors soon after, but moved to find greener pastures in a few years. Sylvania’s poor marketing capability meant the chips he designed, which he thought were good, remained unsold. He took a role at Texas Instruments (TI) in the year 1958. Chang started as an engineering supervisor at TI and soon started managing a full department. He displayed a lot of potential as a leader at TI. His consideration for management positions was contingent on him successfully finishing his PhD.
Chang started his company-sponsored PhD at Stanford and returned to TI in 1964. He picked up where he left off. His stock was on the rise for the next few years and he was soon the VP of the entire semiconductor business. Around that time, TI decided to pivot to the consumer electronics business – calculators, wrist watches and the likes. Chang was chosen to head it.
It wasn’t a roaring success as everyone would have imagined and Chang, self-admittedly, did not understand how to make products that appealed to common folk like you and me. It was too stark a change from manufacturing semiconductors out of pure science and engineering.
Not having the success he had grown accustomed to over the last few years, Chang resigned from TI in 1983 to become the President and Chief Operating Officer of General Instrument Corporation, a smaller peer at the time. General Instrument went through a company-split in the 90s, alongside significant restructuring. Chang did not have to stick around that long to see this happen with his new company.
In 1985, Chang was solicited by Sun Yun-suan to lead a think tank focused on promoting industrial and technological development in Taiwan. Sun Yun-suan is credited as one of the main architects in transforming the island from an agrarian economy to a technology powerhouse in a matter of decades. His profile is an interesting rabbit hole too.
With support from the Taiwanese government and a technical partnership with Philips Electronics in the Netherlands, Chang decided to set up the Taiwanese Semiconductor Manufacturing Company (TSMC). His decision was informed by two things – an identified problem in the market and the constraints of the Taiwanese labour force.
When Chang was with TI, he managed a supplier agreement with IBM, where IBM developed the transistor designs and TI manufactured them. Chang’s involvement in this partnership turned out to be fortuitous. He realised that manufacturing costs of semiconductors were alarmingly prohibitive, left the manufacturers with little flexibility to innovate, and diluted attention between having to build chips and sell end-products that used these chips.
He decided that building a foundry that purely manufactures semiconductors based on customised designs was the way to go. This was further aided by the fact that the Taiwanese workforce lacked the design capabilities or enough wherewithal to build end-products using semiconductors.
This turned out to be a master stroke. With TSMC, there was the possibility for entrepreneurs to start semiconductor companies anywhere in the world. They did not have to worry about upfront costs of setting up a fabrication facility, finding resources for fabbing, or lose sleep over keeping up with the ‘gold standard for semiconductor evolution’ – Moore’s Law (which is awfully hard).
Today’s industry heavyweights like Nvidia, Qualcomm and Broadcom significantly reduced their operating costs when they were getting the company off the ground. These companies, called ‘fabless’ manufacturers had limited options before TSMC came along. They had to go to big-weights like IBM, Fujitsu, Toshiba etc. These massive corporations had special-purpose fabrication capabilities that the fabless manufacturers wanted, but often proposed terms that were untenable. These terms included instances where designs were asked to be transferred as part of manufacturing contracts.
TSMC solved that problem. They also started getting contracts from other semiconductor manufacturers who did not want to use their fabrication capacity for in-demand, but run-of-the-mill chips that could very conveniently be manufactured by TSMC. Chang also did some unconventional things during his time at TI. He priced the semiconductors below the cost-curve to acquire market share. It was not well received at TI although it was left untouched for the time he was there. The experience came very handy in navigating pricing strategies for TSMC in its early days when it had to out-manufacture its competition.
Today, TSMC has an absolute majority of the market share in semiconductor manufacturing. China, despite having invested billions through subsidies and incentives to develop fabrication capabilities, still finds itself a few years behind in its technology. There is just nobody that can compete with TSMC’s capability, especially when it comes to manufacturing high-end semiconductors that are so highly sought after. Samsung is a close second.
The biggest tech companies including Apple, Microsoft, Amazon, Tesla and the likes, design their own chips today. Reasons for this strategic shift are aplenty, but the foremost one is ‘DIY if you can, because it will most likely be better’.
Apple’s M1 chip is a phenomenal breakthrough because of a unified memory architecture that shares a common data pool for its different components. In earlier versions running on Intel chips, each component used different chips and copied data from one another to communicate. They were able to build this beast of a chip having built internal expertise in designing their own chips for iPhones for almost a decade now.
Despite advancements in the capability of tech giants to design chips, manufacturing at-scale is a really tough nut to crack.
Intel announced that it will set up its first standalone foundry for contract manufacturing in the US, earlier this year, and continue to design and manufacture Intel chips. TSMC also announced its intent to set up an advanced foundry in the US, in May last year. Chinese semiconductor fab capabilities are considered to be 5-years behind, but only in comparison with runaway leaders Taiwan and Korea. GlobalFoundries in the US produces about 7% of overall chips globally, but does not compete in the advanced categories and focuses on the simple, yet crucial devices for radio communications, to control display screens, regulate power, or operate tiny motors etc. They do not manufacture high-end chips.
Morris Chang retired from TSMC at the ripe age of 86 in the year 2018. It is safe to say that there aren’t a lot of people that have walked this world and has had this kind of an influence on trajectories of nations and technological leaps.
It is understandable that large economies want to be a part of the semiconductor saga -even if it means a geopolitical song-and-dance.
Continuing in the spirit of leaving you with things that may interest you (presumptuous yes, but you have gotten this far), Google just drastically cut the time that it takes to design semiconductor chips using Artificial Intelligence. Here’s what it means for the market –
Although fabrication of the chips is largely automated, the design still relies on manual processes. Engineers and designers use computer-aided design software, but it can still take them weeks or months to work out how to fit all the components into the available space. Google’s researchers have now shown that the process can be completed in less than a day by using artificial intelligence (AI).
Read more here.
One thought on “When The Chips Are Down, Start a Technology Race”