When can the shortage of chips be alleviated?

I have heard countless times that experts and semiconductor executives say that the semiconductor industry has "finally" got rid of the business cycle. It must have happened dozens of times in the past 50 years. It wasn't true 50 years ago. In the meantime, this has never been true; Even now we are experiencing the latest chip shortage, this is not true.

As long as we use billions of dollars of fabs to make chips, as long as these Fabs take years to build, facilitate and put into operation, cyclical shortages will occur again. Because we simply cannot synchronize the multi-year construction cycle of the wafer factory and the sudden changes in semiconductor demand that may occur in just a few weeks.

The current covid pandemic exacerbated the classic boom and bust cycle of the semiconductor industry, but did not create it.

We ran out of toilet paper due to rush buying! What makes people think semiconductors are different to some extent?

I see that one of the mistakes that the general media often make is that they tend to put all semiconductor production as a whole - usually call all ICs "computer chips". This is not a good representation of how the world makes or uses semiconductors.

Cutting edge components such as microprocessors, mobile phone chipsets and FPGAs are manufactured in the most modern Fabs using the most advanced process nodes - these Fabs cost billions of dollars. Famous semiconductor companies such as AMD, Intel, Samsung and NVIDIA produce the latest generation of chips in these high-end fabs. These companies and their manufacturing partners are building new high-end Fabs at the fastest speed.

Pat Gelsinger of Intel said that Intel is bundling all concrete trucks in several countries / regions just to build Intel's new plants in Oregon, Arizona and Ireland. This is very exaggerated, though not strictly correct. Even so, Intel and TSMC are indeed building many new high-end fabs to try to meet the demand for their most advanced chips.

However, most of the chips we use everyday are not made in high-end fabs. They are manufactured in older Fabs using older process nodes that may have been fully or substantially amortized. This is especially true for parts designed for automotive applications. Cars and trucks are full of microcontrollers, which are not manufactured with EUV or other types of advanced lithography technology, and they do not need them.

The shortage of automotive chips is not caused by the late introduction of 10, 7, 5 or 3 Nanometer process nodes. Car companies saw covid approaching from their rear-view mirror, speculated that covid's influence would reduce new car sales, and then gave up their chip orders.

However, the decline in new car sales proved to be short-lived. Suddenly, the market demand for new cars soared. When automobile manufacturers try to submit new chip orders, they find that the capacity of the Fab dedicated to manufacturing automobile chips has been redistributed to other customers. Delivery times have flashed by, and these automakers now find themselves facing a serious shortage of chips. As a result, automakers began to build new cars and trucks, but they did not complete them because they lacked the semiconductors they needed. No chip, no delivery.

As a result, car and truck manufacturers store most of their finished vehicles where space can be found. For example, it is reported that Ford began to store some finished F-150 pickup trucks in the parking lot near Detroit Airport, the Ministry of public works railway yard near Detroit, and the 30 acre parking lot of Kentucky highway east of Carleton, Kentucky

Ford is not alone. General motors, Ford, Nissan, Daimler, BMW and Renault all announced production cuts in the face of a global chip shortage. Even Toyota, which seemed to be the most able to meet its production needs during the entire covid pandemic, was forced to announce that it would cut its car production by 40% from September this year. Toyota benefited from having a large chip inventory because it revised its business continuity plan after the disaster caused by the Fukushima earthquake and tsunami disaster a decade ago.

However, they did not replenish, in this case, any inventory will not last forever.

The urgent problem here is that automotive semiconductors are not manufactured by special processes such as EUV lithography on the latest 300mm wafer production line. These components are manufactured on older Fab lines, using older process nodes, and typically use 200mm wafers. I thought no one would build more 200mm fabs. They are twenty years old and I thought they were retiring.

But I found that I was wrong from the report entitled "outlook for 200mm wafer plants in 2024" released earlier this year by semi, an industry association serving the global electronic design and manufacturing supply chain. The report said:

"From 2020 to 2024, global semiconductor manufacturers are expected to increase the production capacity of 200mm Fabs by 950000 wafers, or 17%, to reach a record high of 6.6 million wafers per month... During the same period, wafer manufacturers will add 22 new 200mm fabs to help meet the demand of 5g, automotive and Internet of things (IOT) devices for analog, power management and display driver integrated circuits (ICS) , MOSFETs, microcontroller units (MCU) and sensors are in increasing demand. "

However, more research shows that I am also right. According to semi, from 2006 to 2015, the number of 200mm wafers produced per month did decline. The company also said that the number of 200mm Fabs hit the bottom in 2015 and rebounded to 180, lower than the previous peak of 202 fabs. Since then, the number of 200mm Fabs and the monthly capacity to manufacture 200mm wafers have been rising, but not fast enough to meet the recent sudden surge in demand.

The following chart from semi shows the growth of the industry's total monthly production capacity of 200mm wafers and the number of 200mm Fabs expected to go online:

As you can see, the current construction boom of 200mm wafer factories, and the total number of wafers that these factories can produce is also increasing. These figures do not include the high-end 300mm Fabs that seem to have received most media attention.

Therefore, it is not only the capacity of wafer factories that causes the shortage of chips. The figure above shows that the industry is doing well in tracking demand. So, this is another matter. As far as I know, there are two reasons for the current shortage of chips. I think the first and biggest reason is the surge in demand. In the past year and a half, a large part of the world began to work from home. At least for information workers. These employees need a variety of electronic infrastructure to do their work at home, from laptops, tablets and mobile phones to large-scale cloud support, which drives the demand for servers, network and broadband devices and cellular network infrastructure. In turn, the demand for devices drives the greedy demand for chips. Another reason happens to be encapsulation. Liu Deyin of TSMC previously said in an interview that now some people are hoarding chips, which exacerbates the shortage of chip supply.

As far as the shortage in the automotive industry is concerned - well, it seems that it is mainly caused by itself. It is caused by ultra lean inventory control and cruel supply chain management, which need to squeeze every penny from the production cost. On September 19, IDC issued a report entitled "the semiconductor market will grow by 17.3% in 2021 and reach potential overcapacity by 2023", which pointed out that:

"According to IDC data, the industry will see normalization and balance by the middle of 2022. With more large-scale capacity expansion going online at the end of 2022, overcapacity may occur in 2023."

At the code 2021 conference held in Beverly Hills, California, at the end of September, Lisa Su, CEO of AMD, said that it was expected that the chip shortage would be alleviated in the second half of 2022.

Pat Gelsinger, CEO of Intel, predicted that it would take oneortwo years to restore the balance between semiconductor supply and demand. Gelsinger basically has unlimited research resources and can obtain the information needed to support his prediction.

At least compared to the resources available to a single editor, I am happy to use IDC, Lisa Su, and Gelsinger's predictions. Find the balance of forces in 2022, and find the silicon pendulum swinging in another way in 2023.