What’s the deal with Semiconductors, anyway?
Nope, not talking about the pick up trucks... entirely
You may not realize it, but you probably carry a semiconductor around with you at all times. Sometimes even a few of them. Semiconductors are physical substances created to facilitate and organize the flow of electrical currents in devices and equipment. In other, simpler words, they are physical materials that power the microchips in our phones, computers, cars, and more. Let’s put on our engineering hat and dive in to what semiconductors are, how they work, and why they’re critical for our everyday lives.
Because of semiconductors, electronics have become smaller, faster, and more reliable. Made typically of silicon, semiconductors conduct electricity when hot, but insulate when not. Hence, semiconductor, its conductivity (or its ability to transfer electricity or heat) changes based on the environment. These semiconductors are in the heart of computer chips, and known as transistors. Metals are great for passing energy particles around because of the lack of spacing, or gaps, between bands of their substance. Substances like glass, wood, or other non-metallic materials cannot pass energy because of the gaps between them that disable electron movement. Semiconductors are somewhere in-between and enable some passing of electrons.
Transistors, which compose semiconductors, are the building blocks for all electronic devices we use today. Millions of transistors make up these computer chips and that’s how data, commands, and more are passed through computers. If what I’ve been saying hasn’t been nerdy or confusing yet, bare with me as we go further down the rabbit hole. As electricity comes through the gate of a transistor, it gets recognized as a 1. If electricity does not come through, it’s recognized as a 0. This is the value of something being a semiconductor, it creates the binary code that tells all of our devices how to act and respond. Still with me? There are tons of circuits built onto each semiconductor that define all of these binary processes, each composed of millions of transistors, many of which combine to form computer chips.
Semiconductors are freakin’ tiny, that’s how your phone seems like it can do more as it gets thinner. A semiconductor can be as thin as 5 nanometers - that’s the same width of 1/20,000 of a human hair. But the size of it in computer chips also makes it more powerful. The smaller the semiconductor is, the shorter the distance the electricity has to travel, which means better speed and performance. That also means more circuits in the wafer, and lower production costs, and a lower price for the consumer. Cheaper costs and higher speeds are music to the ears of companies like Apple, Samsung, and Huawei - which makes the semiconductor market extremely competitive.
The largest semiconductor companies by revenue include Intel, Taiwan Semiconductor Manufacturing Company (TSMC), Qualcomm, Broadcom, Micron Technology, Texas Instruments, ASE Technology, NVIDIA, STMicroelectronics, and NXP Semiconductors. There are a handful of those companies that you may recognize but unless you’re in tune with the market, most will likely sound foreign.
The semiconductor market has faced some interesting shifts recently that have made huge waves for some of these companies. The first, which may ring a bell, is that Apple has announced it is creating its own chips for its products moving forward. With Apple being one of Intel’s largest buyers for 15 years, that creates a massive revenue problem for the chip provider. Apple wants to vertically own its core technologies, and this will help them drastically reduce costs. On the opposite side of market shifts, NVIDIA has seen huge growth in the last few years thanks to the boom in cryptocurrencies, gaming, and digital-design. Specialists in those areas favor the NVIDIA chip for its capabilities in production and efficiency.
Getting into global geopolitics… there have also been major concerns with TSMC controlling around 50% of the chip market due to tensions between Taiwan and China, and China and the US. TSMC are industry leaders and used for hundreds of thousands of electronic devices. This thread between Josh Wolfe of Lux Capital and Adam Bain at 01Advisors (and former COO Twitter) sheds some light on just how volatile the market can be and why there is potential for some major, major issues.
You may have seen the news recently that “there is a major chip shortage” and ignored it, like I did, because you didn’t understand it. Here’s Morning Brew’s brief overview of the issue:
The chip shortage is impacting everyone - from car companies like GM and Ford, to electronics manufacturers like Dell and HP, to gamers trying to build their computers. It has radically displayed systemic supply chain issues related to the semi industry, and puts the US in a precarious position of dependency on suppliers from abroad. This is an industry approaching half a *TRILLION* dollars in size, and the US is freaking out about losing a significant portion of that to these overseas suppliers as we become more and more reliant on computers in our everyday lives.
So where is the semi market headed, and why is the competition being so quickly drawn out? Looking at the landscape of where technology is headed there are a handful of drivers for the future. Artificial intelligence, IoT, AR/VR, gaming, blockchain/crypto, and all the other future technologies will be strongly reliant on what chipmakers can produce, and we will almost certainly see another boom in the volume of chip companies out there.
A frequent mention of mine in these newsletters (for good reason), Chris Seifel, has several great pieces on where the chip market is headed - and the focus on the AI market is the most intriguing to me. A key point that he made is that there is a “critical role that computing processing power plays in [AI’s] development. AI chips are known as AI accelerators, and they are built to optimize artificial neural networks (ANNs). An ANN is a computing system that looks something like a human brain and is meant to process and analyze data the way that we would. ANNs have hundreds of thousands of artificial neurons, called processing units, that are composed of input and output units. Input units receive structures of data, which then gets learned by the neural network, and then gets pushed through the output as a different report - similar to humans learning rules. ANNs get “trained” to recognize patterns in data, and when the actual output does not match the desired output, it is adjusted using backpropogation. Here’s a real-world example:
During the training and supervisory stage, the ANN is taught what to look for and what its output should be, using yes/no question types with binary numbers. For example, a bank that wants to detect credit card fraud on time may have four input units fed with these questions: (1) Is the transaction in a different country from the user’s resident country? (2) Is the website the card is being used at affiliated with companies or countries on the bank’s watch list? (3) Is the transaction amount larger than $2,000? (4) Is the name on the transaction bill the same as the name of the cardholder?
The bank wants the “fraud detected” responses to be Yes Yes Yes No, which in binary format would be 1 1 1 0. If the network’s actual output is 1 0 1 0, it adjusts its results until it delivers an output that coincides with 1 1 1 0. After training, the computer system can alert the bank of pending fraudulent transactions, saving the bank lots of money.
What this means is that the AI chip market is going to have immense demand as AI starts to become even more common in our everyday lives. Allied Market Research believes that the AI chip market was valued at $6.6B in 2018 and will grow to $91B by 2025, one of the largest shares of the entire semiconductor industry. With the proliferation of technologies, we’ll also see a major impact to the need for edge networks, like Cloudfare’s. More devices online and more computer processing power means a need for more bandwidth and speed.
While I hate to share the same conclusions as others - at least after quoting some of their research in here - Seifel and I are on the exact same page on the fact that we will see rapid growth in the use case-specific applications of semis. More and more semi manufacturers will pop up whose expertise lies in edge network semi manufacturing, AI semi manufacturing, IoT-specific semi manufacturing, and more. These companies likely will outperform some of the legacy manufacturers on application specific (ASIC) tasks, but I think we’ll see an incumbent acquisition spree paired with the proliferation of start-ups. There’s lots of capital to play with and an ever-growing market size.