by 
With Intel Corp. predicting a scarcity of semiconductors that will last years, chip makers are set to build 29 new fabs by the end of 2022, according to industry group SEMI. These plans will have little impact on the current shortage, which has spread well beyond the automotive industry.
Although fab plans represent a boon for equipment manufacturers, it doesn't mean an influx of chips will be hitting the market. Semiconductor lead times have increased to 18 weeks, according to Bloomberg -- the same period required to manufacture a chip. Experts say the lead time expansion indicates chip makers' ability to meet demand is worsening.
According to LevaData, between January and April 2021, semiconductor lead times increased 75 percent on average. For some components, increases have been steeper:
- Programmable logic: 12 to 33 weeks (175 percent increase)
- Microcontrollers: 16 to 44 weeks (175 percent increase)
- Network interface ICs: 12 to 40 weeks (233 percent increase)
- Serial IO controllers: 9.5 to 39 weeks (311 percent increase)
Fabs, fabs, fabs
In January -- when the semiconductor shortage first became apparent -- experts said chip companies should reconsider their long-term approach to capex and capacity.
“Generally speaking, chip companies are building to demand and available capacity,” said Syed Alam, semiconductor global lead and managing director for business consultancy Accenture. “Demand for semiconductors is not going down. More applications are using more chips – for example, the chip industry was driven by PCs, then mobile and consumer products. Now there are automotive, AR/VR and the cloud. All of these applications will use chips. [The industry] should not consider this the shortage of 2021 – this could happen again, or it could last for a longer period of time.”
It appears the industry listened. Globalfoundries broke ground this week on a new 300-mm fab in Singapore, one of a projected 19 new high-volume fabs to be built by the end of this year, according to SEMI. Another 10 are expected to be built in 2022.
Taiwan Semiconductor Manufacturing Co. (TSMC) and other Taiwanese chip manufacturers are projected to construct six new fabs in the coming year, with two more coming online in 2022. TSMC has said it plans to invest about $100 billion over the next three years to increase foundry capacity.
Chinese manufacturers are expected to keep pace over the next 18 months, with a half-dozen new North American fabs planned by the end of next year.
Fabs that will produce 300-mm wafers account for most of the new capacity, SEMI said. No less than 15 300-mm lines are under construction, with seven more planned in 2022. SEMI said those seven will consist of a mix of 200-, 150- and 100-mm production lines geared toward meeting unrelenting demand of automotive, 5G and Internet of Things components.
Fifteen of the 29 new fabs projected by SEMI will produce upwards of 30,000 wafers monthly. A key driver is the memory sector, that will account for at least four dedicated fabs over the next 18 months.
While foundry construction is well underway, SEMI does not expect chip makers to begin installing lithography and other equipment until 2023 “since it takes up to two years after ground is broken to reach that phase.”
The first problem
Of course, the current dilemma began with forecasting, which no amount of fabs will fix. OEMs scaled back their chip demand late in 2020 anticipating a soft market due to the Covid-19 pandemic. Instead, demand for consumer and home electronics went through the roof at the same time people were planning to buy new cars.
It's clear that businesses recognize the problem. The electronics industry is pushing for visibility -- more information from more tiers of suppliers. Materials or substrate shortages, for example, can significantly impact component production schedules.
Visibility is driving the demand for digitization, which will accelerate supply chain data collection and analysis. Communicating delays in real time allows companies to respond more quickly.
Additionally, the automotive and chip industries have agreed to work together to improve forecasting. SEMI and the Center for Automotive Research (CAR) have signed a memorandum of understanding to advance collaboration between the two industries. The MOU lays the foundation to connect microelectronics manufacturing and design stakeholders with the automotive and mobility ecosystems through programs and events that advance both industries.
The automotive industry, reports KPMG, could lose $100 billion in 2021.
Some consultancies are recommending the supply chain carry more inventory. Just-in-time (JIT), build-to-order (BTO) and lean have reduced the number of parts in the pipeline, leaving little wiggle room for upside demand.
But the production of components is still heavily tied to forecasting. Twenty years ago the chip industry faced a very different problem: the unbridled optimism of the dotcom boom fueled an oversupply of parts. Semiconductor companies wrote off, or wrote down, $13 billion worth of inventory.
JIT, BTO and lean were among measures implemented to avoid a similar excess. Now, companies are looking for ways to mitigate shortages. Building fabs is part of the answer, but unless OEMs and the supply chain can improve the accuracy of their forecasts, the chip industry's next problem could be be overcapacity.
