Why are we so dependent on Taiwan for microchips? Why couldn’t these be manufactured elsewhere, in addition to Taiwan?
In: 1601
Sure, they could. But the company that makes them, the Taiwan Semiconductor Manufacturing Company (TSMC), has some unique business benefits from the Taiwan government. Other locales have offered similar incentives, but the government of Taiwan sees TSMC as a strategic shield against a Chinese invasion of the island. They are strongly incentivized to have things go “very bad” should China invade them, given how many times Chinese leaders have proposed invasion in the past.
They are manufactured elsewhere, TSMC is just the biggest company. Chip foundries are extremely specialized and require a lot of investment into equipment and facilities. Companies are working on building more to keep up with demand, but that process takes several years. There is just too much demand for the handful of chip foundries around the world.
We’re not. Taiwain Semiconductor Manufacturing Company has fabrication sites in various countries, including the PRC, Singapore, and the US, although the bulk of their facilities are in Taiwan itself.
It’s just very expensive to spin up a new fabricator from scratch, so the response to a shortage might take years to manifest as a new fabricator.
TSMC (Taiwan Semiconductor Manufacturing Company, Limited) is the largest microchip manufacturing and design company in the world, controlling 51% of the microchip market. The second and third largest producers are Samsung and Intel.
Only these three companies can even produce today’s most advanced microchips. Microchip production relies upon extremely expensive factories and foundries. TSMC’s latest factory cost $19.5 billion.
So TSMC simply became the biggest player in the industry, with the massive capital required to manufacture semiconductors. TSMC is also majority-owned by foreign investors, the top ten of which are American investment firms. The company started in 1987 as a collaboration between the government of Taiwan, the tech giant Philips, as well as private investors with an interest in semiconductor technology. It’s stayed at the forefront of the industry since.
The problem isn’t semiconductors in general but the highest technology factories. There are dozens of fabs located worldwide that can build prior generation technologies. The main issue is that there are probably only three companies today that can make very high end products like the latest CPUs – Samsung, TSMC and Intel. Intel is falling behind the curve a bit and Samsung isn’t as focused on CPUs. TSMC is pretty much the only game in town for the latest AMD and perhaps soon to be Intel CPUs and the latest GPUs (likely). Unfortunately without CPUs, the other chips (even though much higher volumes) aren’t too useful.
These high end factories takes dozens of collaborations across the industry, multi-year partnerships with customers, a stable well trained workforce and tens of billions of dollars to get to work. No other company (at the moment) can duplicate this without many years (think decades!) of development and investment. TSMC has been in the business for nearly 40 years and their investment isn’t going to be duplicated any time soon.
Even if other companies decide to invest in current gen technologies (unlikely), it would be useless for companies who are designing for the next generation technology. It is even less likely that any customer would bet their next gen products on a company that doesn’t have a solid track record.
One major factor is that the materials to manufacture them are available in abundance in the area, so that means less transportation cost for raw materials, cheaper materials, and less issues sourcing them. Next is cheaper labor, and finally once they have them manufactured they are on one of the largest major international shipping centers, so they can be transported easily to where needed.
Other places DO manufacture them, but it isnt something anyone can do. The infrastructure needed to manufacture them is specialized requiring special equipment and that costs, then training, then all the things it takes to get their first product out is a 4-7 year process. Taiwan’s government subsidies their chip manufacturer to keep them vital to the world market to deter invasion from a certain country that likes to rattle the saber far too often. TSMC’s largest plant was reported to cost over 20 billion to get from start to producing. Not many people can invest that much to compete, and those that do know the time it takes to get up isnt a fast process.
THEN you have to have complex trade agreements, patents, and processing deals to make that investment worthwhile, and those take a lot of time and money to get going too. Then after all that, the next gen chips start being produced to be old tech requiring new tooling and equipment each generation upgrade.
To be clear, the Mikrochips could be produced somewhere else, but the process is not easy and involves alot of very expensive and very precise machines inside a Clean room.
Operating and build auch big clean rooms is the first expensive steps. Following buying and building up those big really expensive machines, talking about the sice of a double bed, two meters high, and not only one of those, but alot of different ones for different processes.
All those machines have to be fine tuned to make really precise structures that are needed, so you need qualified people to build the machines, maintain the machines and use the machines.
Then the whole procedure, depending on product, might take weeks or month to produce. So there is no quick way to check if a machine is working.
Working in a semiconductor fab, the fastest priority (hand delivery inside fab plus the notice “drop everything and do this now!”) still took about 3-4 weeks to be completed.
So that’s already alot of work done before you can even start mass producing.
And if you want to take jobs you need to show some quality or capacity first.
So getting to run a fab for Mikrochips is the really hard part here
Imagine you want to open a woodworking shop. Strictly speaking, all you need is a few basic tools, like planes, saws, measuring instruments and so on, and a moderate amount of training for someone to begin making simple things. Good to go, right? We’ve been doing woodwork for tens of thousands of years, in some capacity.
Now, maybe woodwork isn’t enough for everything. So, maybe you turn to metal work. Cool, right? The only issue is, you need more specialized training as metal is more difficult to work with, and you need tougher tools, especially to handle something like stainless steel. If there’s one million places that could make a woodworking shop, there’s maybe a hundred thousand that could open a metalworking shop.
Now, metalwork’s awesome! You can do tons of stuff with it, from vehicles, to decorations, to analogue computers, but as a society gets more and more advances you need more advanced ways to track things, do accounting, process word documents and perform simulations, so you probably need a computer, and that’s where it gets tricky.
A lot of the parts of a computer are pretty easy to make. Cooling systems are pretty much universal, PCBs are also pretty easy to make in the grand scheme of things, and cases aren’t bad, either. The issue is the silicon.
Early silicon nodes were pretty easy to start with, but as they got more and more specialized we started needing more and more special tools over time, and many people in that market started dropping out because it just took too much funding. In contrast, because economic independence is very important for Taiwan, the government invested very heavily into TSMC, so that they could making new “nodes” or “ways of making silicon smaller and faster” every year for decades, until it was basically just TSMC and Intel left, at this point, and even Intel is still kind of behind.
If one million places could make a woodworking shop, and one hundred thousand could make a metalworking shop, and ten thousand could make an early silicon node, well, maybe only ten or so could make a high end silicon manufacturing plant nowadays. And even then, if you choose to make one, they take probably five or so years to really get into working order, and it takes two or three generations before people will trust you as a supplier of silicon, so it really takes about 10 years to react to the market in this space.
We’ve recently had some really big changes in silicon demand in the last two years, and I suspect in the next eight we’ll see a different landscape in terms of who is offering silicon nodes for high quality semiconductors, as many players in the space are able to react to recent shortages.
A Taiwanese here. The Taiwan Semiconductor Company (TSMC) is in charge of fabricating a lot of the world’s chips, even though they have plants across the world. The Taiwanese government and people are very invested in its success. Therefore, some of the most talented Taiwanese people go work for the company in various capacities and the government provides lots of research help, which allows it continually improve its products. While lots of products do not necessarily need the most advanced chips, it’s reputation and scale has made the company a key supplier for the material.
One of the things that moved manufacturing overseas that I’m not seeing mentioned elsewhere is the waste products generated by chip fabrication, and the regulations around disposal of same that exist here in the US that are more lenient or non-existent elsewhere.
The waste products are very dangerous to be around, handle, and transport, and nobody wants to take them and store them state-side.
Latest Answers