Why will there be a shortage of automotive semiconductor chips in the next 1-2 years?

Roland Berger's recent article "Steering through the semiconductor crisis A sustained structural disruption requires strategic responses by the automotive industry" is really interesting. The description in the article is somewhat similar to what we are seeing now.

Roland Berger's main views are:

●The global semiconductor shortage is expected to continue beyond 2022, and the chip crisis in the automotive industry and many other industries will last for several years - the decline in automobile sales in 2019 made the chip supply appear sufficient, but the sharp contraction in the first half of 2020 and the retaliatory rebound in the later period cleared the overall chip inventory, and the impact is expected to continue for 1-2 years.

●From 2020 to 2022, chip demand will grow at an annual rate of 17% (mainly starting from Q4 2020), while chip supply will only increase by 6% each year.

●The main reason for the shortage of automotive chips: The biggest shortage occurs in the old generation of chips (MCU and power chips in ECU under distributed architecture) and traditional semiconductors that traditional internal combustion engine vehicles rely on. However, the current new production capacity is mainly built for the new generation of centralized architecture (Domain or even Zone architecture). From the actual results, the new investment will not alleviate the original shortage.

▲Figure 1. Chip supply situation (matching of demand and supply)

Part 1 Mismatch between chip supply and demand

From a technical perspective, the distributed vehicle architecture of most vehicle manufacturers currently relies on ECU technology. This system is composed of a large number of microcontrollers (MCUs), each of which has limited computing power. The automotive industry accounts for approximately 40% of the demand for MCUs.

In other words, the increased demand for purchasing MCUs in the automotive industry in recent years is not in sync with the demand in other industries, and chip manufacturers have no motivation to expand the production capacity of these old processes.

According to Roland Berger's research, consumer electronic devices also need to increasingly adopt chips using older technology (40-90nm process). Of course, these production capacities are not used to manufacture MCUs, but for 3D audio, fast charging and 5G, which require power chips, RF and audio semiconductors.

Therefore, in the process of limited production capacity and having to rely on grabbing resources, consumer electronics manufacturers, because they are quick to pay and lower prices (they have higher purchasing power and bargaining power), will further amplify the shortage of automotive and industrial chips.

▲Figure 2. Semiconductor manufacturing process of different products

Looking ahead, automotive chip companies have also discovered a most practical problem: the automotive industry is undergoing a transformation towards smart cars and transitioning to new electronic architectures, such as Domain and Zone architectures. The new demand for chips is mainly for high-computing heterogeneous computing platforms to replace the original ECU. Automotive chip manufacturers will pay close attention to this transfer and change. Around automatic assisted driving technology, infotainment systems and power system management, OEMs will directly contact chip suppliers to discuss future cooperation, and naturally figure out which are temporary needs and which are long-term needs.

This kind of involution is mainly caused by new car companies, which have kept a certain distance from the design concept of distributed architecture from the beginning, so they can quickly migrate to a centralized architecture (mainly due to organizational structure issues). Since they have been in the distributed architecture for a relatively short time, they are relatively less affected by chip supply, which also forces traditional OEMs to come up with extra motivation to transition to a centralized architecture.

Note: At present, the three-electric cross-domain architecture will appear around Q3 2022, and the first generation of architecture with a central computing platform will be ready for SOP by the end of 2023 or the beginning of 2024. It is really competitive.

Figure 3. Different electrical and electronic architectures

When everyone started to switch from the "just-in-time" to the "just-in-case" method to purchase chips, it objectively caused order congestion and made the market-based regulation mechanism of orders ineffective, which will inevitably cause huge fluctuations in potential demand and sluggish demand in the future.

I think the whole logic is this: in the entire chip industry, due to the continuous advancement of consumer electronics and IT technology, investment in chip manufacturing capacity usually focuses on expanding cutting-edge manufacturing capabilities. From the perspective of giants such as TSMC and Samsung (and their younger brothers behind them), investing in advanced semiconductor process capacity can maximize the service life of assets, protect billions of dollars in investment, and optimize and reduce long-term manufacturing costs.

The compound annual growth rate of advanced manufacturing processes from 2020 to 2022 is 26%, while the process capacity required for automotive electronics is only about 2% per year. This objectively slaps the automotive industry in the face - insisting on its own slow pace and thus being unwelcome in the chip industry.

▲Figure 4. Overall chip process changes

There are also all kinds of strange emergencies that make semiconductor supply chain disruptions more and more frequent. For example, various climate changes (more dangerous tropical storms in Southeast Asia, the semiconductor manufacturing center, and the impact of severe droughts in Taiwan on water-intensive semi-finished products) and epidemic factors, as well as rising political uncertainty, have an increasing impact on semiconductor manufacturing.

▲Figure 5. This world is not a smooth sailing world

Part 2 Chip Alternatives for Chinese Automobiles

Similar to the problems in the automotive industry around the world, many new chip cars have emerged in China, which can be described as a thousand ships competing in a race.

Many companies start from a single product category, and many of them rely on processes above 40nm to meet the reliability and stability requirements of automotive chips. The production capacity of these automotive chips is limited. Even more cruel is that China's new car companies and joint ventures are more inward-looking. Except for a few chassis, airbag MCUs and safety parts, many functions are being quickly integrated.

We can calculate the following data:

Which ECUs are irreplaceable in the short term? How many are there and how much room is there for replacement?

●How fast and how deep will the OEMs move towards the Zonal architecture? With one or two companies running fast, how fast will the rest of them move?

▲Figure 6. Xpeng's G9 successor EEA3.0 will start delivery in Q3 2022

Figure 7. The EEA architecture of traditional automobile companies is also changing rapidly

summary

So in the final analysis, all automobile companies are trying to make smart cars and new things with new architectures. However, no matter how much they sell, there will still be a gap in chip supply.