Made in Taiwan. We are tired of seeing this phrase in almost all products, especially electronics, that pass through our hands. It is probably a country that you will hear very little about, or perhaps you will confuse it with its neighbor Thailand, and in fact, there are people who think that it is part of China, but in reality, it has enormous importance and is that the current and future innovations of the world They depend on one country, whether we like it or not: Taiwan.
We live in the age of technology, where chips and processors are literally present even in refrigerators. And the semiconductor industry, in which these chips are made, is on which all the innovations in this regard depend. Surely if we talk to you about TSMC the name at least sounds familiar to you and if it doesn't it will start to sound like it since in this regard everyone depends on a TSMC and will continue to do so for at least the next decade.
The world wants (and needs) better semiconductors
Semiconductors have and have had a significant impact on our society. They are everywhere and are a fundamental part of any electrical object, from a simple LED to a shuttle that can send humanity to Mars. Without them, you couldn't be reading this article, we couldn't drive cars, and you couldn't call your family on the phone. They are necessary to build an electronic device, from televisions and computers to medical equipment, going through everything you can think of.
We can illustrate this fact with the vaccine for COVID-19, created in less than six months from the beginning of the research to its mass production. This would have been impossible a decade ago, as video conferencing can now be held to connect researchers from all over the world, and all individual work from each team is instantly available to everyone else. Machine Learning was used to sequence the DNA of the virus, state-of-the-art machines were used to prototype potential new vaccines, thousands of volunteers were instantly contacted to test the vaccines, and then millions of doses were manufactured on a large scale, all of it. in six months. This process would have taken years a decade ago.
Doing these little things may seem easy, but making them better, more efficient, and smaller is the real challenge. Without the progress the semiconductor industry has made, we would not have been where the world is today, and even if the software or larger hardware is the final means, semiconductors do the final work to make them work.
So it is clear that we will need better semiconductors in the future. We want electronic devices to consume less electricity to save the planet, we want clean energy to replace fossil energy, we want unlimited battery life for our smartphone or laptop, we want a smaller camera but with a higher resolution, and of course we want better computers to run AI, we want to travel to space, we want… everything. More, better, and faster.
TSMC and Taiwan, essential and vital for progress
To understand why TSMC and its namesakes are so essential, we must first understand the supply chain of any electronic device used by the end-user.
Historically, all electronic hardware sold goes through a small business unit that makes the integrated circuits (ICs or chips). It costs time and money to create them, and yet these companies did not want to waste time manufacturing these chips, so they ask large companies like Texas Instruments or Samsung to help them do it. But those don't necessarily have the ability or the will to make chips for the competition, and if so they often ask for a lot of money to make them.
It is then that we begin to see the birth of a company focused only on semiconductors like TSMC, which is exclusively dedicated to designing and manufacturing them. Some of these companies only design them, and they are known as Fabless companies (without factories), which only have to make orders to the producers saying what they want. TSMC is a bit different as they only make those chips, but it is another company that has to give them the design they want, and this is known by the term 'Pure Play Foundry'.
To better visualize and understand who is who in this supply chain, let's name the actors:
- End product company: Amazon (cloud servers), SONY (consoles), Apple (smartphones), Tesla (cars), etc.
- IDMS (who design and sometimes produce the chips): Samsung, Intel, etc.
- Fabless (which only design chips): NVIDIA, AMD, Qualcomm, Broadcom, etc.
- Pure-Play Foundry (only makes the chips): TSMC, GlobalFoundries, etc.
Designing a chip has its own peculiarities and cost, and in the following graphic, you can see that it can cost more than 500 million dollars just to design one.
As we have already said, TSMC only manufactures the chips and avoids the cost of the design, so the manufacturing must be simple because it is only following a design, right? Actually, the most complicated part is being able to translate a design into a material chip, and in fact today only two companies in the world do it: TSMC and Samsung.
Why is it so difficult to make a chip?
In order to make smaller and smaller chips, TSMC uses an extremely complicated process called extreme ultraviolet lithography, or EUV. The machine that enables this lithography process is provided by a company called ASML, and what this machine does is basically a new way of producing light. The process involves dripping molten tin into a chamber and hitting it with a laser to create a glowing light/plasma that can pass through the series of mirrors and masks to the wafers where the chips are made, and this happens about 50,000 times per minute.
Some agree that this machine could be probably the most complicated mankind has ever created, even more, difficult than putting a man on the moon. This type of machine alone costs about $ 120 million, weighs 180 tons and is the size of a double-decker bus. A curious fact is that ASML needs a whopping 5,000 suppliers to manufacture just one of these machines.
Some of you will think that it would not be enough to buy the machine from ASML so that anyone could make their own chips? It is so simple actually since to make the chips not only with this machine is enough, but there are other equipment that comes into play and, in fact, the machine only does around 10% of the process. TSMC needs to learn to operate each piece of equipment together, and each piece of equipment has its own quirks, faults, and tolerances. Having to chain all the equipment together, in a high-volume manufacturing process, is already a problem in itself.
Bottom line: ASML sells the pot and TSMC sells the complete cooker. Cooking is very difficult, since each one takes a lifetime of specialization for their favorite dishes, and it is difficult to be really good. For more INRI, all the work is not done by a single operator but by a whole army of people who have to work together. Also, the people who make chef's knives aren't necessarily good chefs, right? The machines that ASML builds are simply a tool for solving a problem, and a good hammer needs a good craftsman to handle it.
Minimize failures, research, and develop
Getting as few defects as possible is the biggest challenge for any manufacturer. This is probably the main focus for chipmakers, and where TSMC stands out above the rest in its Taiwanese factories. The percentage of chips that work properly is called yield; If you manufacture chips but more than 60% are defective, you would not really be very profitable, and that is when you have to find the reasons for the failures in order to eliminate or reduce them.
Performance is why manufacturers spend so much money on research and development. They always want to get the highest possible performance, to be able to scale to billions of chips to satisfy their customers and to realize greater profits. Every year TSMC customers ask for better and smaller chips (especially Apple), and the process to make them needs to be refined, and they constantly need new equipment to support these processes. In 2019 alone, TSMC invested more than $ 3 billion in R&D, what other company would have the budget to do this?
That is why we often speak of "mature nodes", since they have been in operation for a long time and experience has allowed them to fix bugs, invest in development, and significantly improve performance. TSMC's current monopoly is probably the result of 40 years of experience in it.
Conclusion: why will Taiwan dominate all innovations?
We have been talking throughout this article about TSMC, a company whose headquarters are in Hsinchu, Taiwan, and in 2013 it already had more than 30,000 employees. It is a very high capital company and one of the few that can afford to invest billions in development every year, so it is not surprising that most Fabless companies decide to commission TSMC to build their chips, already despite the fact that when changing nodes the performance is usually low, the reality is that investment in research and development allows them to increase this performance quickly.
Right now, and as we explained at the beginning, practically anything that occurs to us related to development and innovations has to do with chips and processors, chips and processors that mostly pass through our hands at one time or another. from TSMC, a company based in Taiwan. And it is in Taiwan where all this development takes place, and therefore we depend on Taiwan for it. Whether we like it or not, Taiwan is the country that dominates and will dominate all technological innovations globally.
And why not make "another TSMC"?
Because if it were simple they would have already done it, since nobody likes to depend on a monopoly, even if it turned out like that and it is not something they have sought as such. Every year, billions of dollars are invested in R&D to maintain its advantages, and just building a factory like the ones that TSMC has in Taiwan would cost precisely billions:
- TSMC's last fabrication plant (Fab18) to make 5nm chips with the new technology cost $ 17 billion, excluding R&D.
- TSMC's next factory for 3nm chips costs $ 19.6 billion.
And in the latter case, they estimate that it will take about 5 years to start making benefits and that in the best of cases. Companies should thoroughly research product demand, including possible long-term changes, before beginning the construction of factories (and this, with the contracts they currently have, is not a problem for TSMC as they have more demand than offer).
The most serious competitor that TSMC currently has is probably Samsung, but as we have explained on other occasions, there are quite a few problems both due to legal or intellectual property issues and production capacity, something that we have already seen with the NVIDIA Ampere chips made by the Koreans.
But not everything is money or patents, when we talk about semiconductors we must also talk about physics. For example, Intel is not having problems with its nodes because they lack time or money, they simply lack the experts that TSMC has on their staff. The number of engineers who have the knowledge of how to assemble the chips is very small, negligible in fact, compared to the software or hardware engineers in the world. The problem is that all great companies want the best, but there are not many graduates in this type of job and, even then, they need many years of training in the company itself. The barriers to entry are extremely high when it comes to chip making.
An example is found at Apple, which is well aware of this problem and has certainly already considered having its own manufacturing facilities. With 200,000 million in the bank, it is easy to think about it, but in the end, it all came to nothing because they were never sure that they would be able to manufacture chips with the quality and efficiency that TSMC has in Taiwan. Some Chinese manufacturers are trying, but for now, they are light-years away. That said, like it or not, we depend on Taiwan for development and innovation.
0 Comments