This post is an excerpt from our newsletter in which we take a deep look at the shifting trends of the semiconductor industry. If you would like a full copy of the note, please contact us directly.
The market for semiconductors has changed rapidly in recent years, and these changes will have a major impact on the future of electronics, technology and likely the whole Global Economy. In this note, we are going to look at a key development in this changing landscape – the shifting economics of producing custom ships. The number of non-chip companies designing their chips has grown markedly, becoming a major competitive advantage in many fields. At the same time, the costs of design has fallen notably as a number of third parties have emerged to facilitate the design flow. However, the chips are not truly ‘cheap’, and the ways that companies are choosing their designs tell us a lot about where their priorities lay. We will examine these changes and walk through the universe of companies at play in this space.
First some background on semiconductors. If you are already familiar with this skip ahead a page, but maybe bookmark this section the next time you need to explain to your in-laws what you do.
Semiconductors, aka chips or semis, are highly miniaturized electronics that do a lot of math very quickly, and through this math they can make things happen in the real world. To put it very un-elegantly, chips are the brains that power our electronic devices. They help computers and other machines evaluate alternatives, powering phones, computers, cars, airplanes, the Internet, etcetra.
Semiconductors are very sophisticated objects that are manufactured on silicon wafers. The manufacturing for these wafers is very expensive, requiring billions of dollars in upfront investment. One of the great technological miracles of the modern age has been the way the chip manufacturers have been able to steadily shrink the size of the transistors populate chips, meaning that the economics of chips have been improving steadily for 40 years. This is the phenomenon known as Moore’s Law. Advances in Moore’s Law are denominated by the size at which their smallest circuits can be built. Today the smallest process is 7nm, which is very, very small.
The economics of semis are a function of applied geometry. These wafers are circular and so the challenge is to cram as many individual chips into each wafer as possible. Adding a feature to a chip makes the chip larger and so fewer fit on a wafer. The design of a chip is thus a trade-off between functionality and cost. Fortunately, the price that customers pay for each chip is a function of market economics. Build a better chip and people will pay far more for it than it cost to produce. Put simply, the cost of chips is determined by geometry, while the price paid for those chips is determined by the market.
The manufacturing lines for semis, which we call fabs, are very expensive. As a result, today only a handful of companies can afford to build chips using the most advanced techniques. As a result, the industry has adopted a two-part model. There are companies that design chips who then pay other companies to produce their chips on their fabs. These third-party fabs are called foundries. The design companies build what is essentially a blueprint for the chip, and then the foundries do the physical manufacturing. Bridging these two is thus a major source of friction for the industry, and, as we will discuss below, is a source for some of the changes we are seeing.
Also, when looking at semiconductors it is sometimes useful to divide the products into two categories – analog and digital. Digital chips do work entirely in the realm of the digital, they compute solely based on digital inputs. By contrast, analog chips have some interface with the physical world – processing radio waves or sound, for example. This distinction is somewhat blurry as analog chips end up doing a lot of digital processing. However, from a market viewpoint there is a clear distinction typically based on the business of the chip’s end-customer.
Lastly, let’s provide some examples to make this more tangible. The most common type of digital chips are CPUs (Central Processing Units) which are the key component powering computers. Mobile phones also have something analogous called an Applications Processor which is what makes all the apps run on a phone. Analog chips come in many varieties. A very common one is the Baseband of a mobile phone which converts the signals from a radio wave into information like voice calls and all the data needed for your apps. (Technically the baseband is a digital part, but it sits on top of a series of other chips that directly interface with radio waves.) Other analog chips include sensors in cars to measure temperature and tire pressure.
The semiconductor industry as we know it today grew out of the electronics industry over the past 50 years. Originally, most electronics companies made their own key components. As these components become more advanced, semiconductors emerged as a product category of their own. However, that legacy persisted with many electronics companies manufacturing their own chips. About twenty years ago the cost of building semiconductor fabs grew so high that the split we have today between Design companies Foundries started to emerge. Many companies selling into analog markets still own these older fabs, which works perfectly well in many cases. However, for companies operating at the edge of requirements – the fastest, smallest, most powerful or power efficient chips – needed to make use of the most advanced manufacturing technology.
Today there are just three companies even attempting 7nm – Taiwan Semiconductor (TSMC), Intel (who is a bit stuck) and Samsung who is likely close. A fourth, Global Foundries, recently announced that they will stop at the previous node – 10nm. (For more on this see our post). By contrast, there are hundreds of design companies, as well as dozens of analog companies that do both design and some of their own manufacturing.
However, here we are also seeing a wave of consolidation. Which brings us to our first fundamental reality of the industry – the easy days of growth are behind us. For the past twenty years, the industry has enjoyed a steady series of demand waves. First PCs, then Internet connectivity (which greatly drove demand for computers), then TVs and game consoles and most recently mobile phones and then smartphones and cloud computing data centers. Those product categories are now fully saturated. Demand continues to grow, but at much lower rates. New product categories like AR, VR and autonomous cars still hold promise, but the math makes it very hard for a $483 billion industry to grow at double digits.
As a result, companies are starting to realize that the only way to grow is to acquire competitors. This transition is still underway, and many industry veterans are only now coming to grips that the semis industry is no longer the ‘hot’ growth segment they grew up in, but more closely resembles the steel or automotive industry. Ten years ago there were 500 semiconductor companies, today there are something like 150. And in ten years there will probably be fewer than 50, or fewer.
This consolidation is good for the industry from an economic viewpoint. It had to happen, but that is of little consolation to customers. In many product categories we see effective duopolies or triopolies. Larger customers no longer have the pricing leverage they once enjoyed, and smaller customers frequently complain about how hard it is for them to get attention from their dwindling pool of suppliers. Again, from an economic viewpoint, this is just the market finding equilibrium, but it is also opening some new doors.
As we noted, once upon a time, all electronics companies built their own chips in-house. As the costs of staying in the business rose, they gradually offloaded their design teams to work with merchant suppliers, companies who sell to any customer. The merchant vendors were able to amortize their R&D costs across multiple customers which gave them a meaningful cost advantage over the in-house teams, further extending the cycle. However, with the gradual erosion of pricing power, this math has shifted slightly. So while chip companies are enjoying a ‘golden age’ of pricing power, customers are looking for alternatives.
This brings us to an emerging trend of non-chip companies once again building their own chips. Today we see the leading technology companies – Apple, Amazon, and Google, among others – building internal chip teams.