The last 6 months have brought a raft of announcements to increase investment from major semiconductor manufacturers, especially foundries that provide wafer processing services to fabless semiconductor companies. As part of ‘IDM 2.0’, Intel announced intentions to invest $20 billion on a new foundry services venture. TSMC increased it’s 2021 capex guide and estimated it would spend $100 billion over the next three years. And most recently, Samsung raised the estimate for non-memory semiconductor investment to $151 billion by 2030…
Yole Développement delivers its technical & market expertise, through a dedicated article today. Its aim is to point out the status of the semiconductor industry and the current semiconductor shortage, largely covered by media and presenting lot of questionings coming from the semiconductor players. This snapshot, based on both quarterly market monitors, Application Processor and Microcontroller as well as on the large collection of computing technology & market reports, underlines largely the key decisions announced by leading companies in term of investments.
These are towering investment estimates, well above what industry observers would have thought just one year ago. How did we reach this point? Let’s look back at the events and circumstances that brought us here.
Many factors are reshaping the foundry landscape; some of these have arisen suddenly while others have taken considerable time to develop. Influences endemic to the industry itself and those coming from broader geographical and political factors are playing major roles today.
First you have a natural byproduct of the state of the industry’s technological advancement. To reach the highest level of process technology requires massive investment in R&D as well as ever-larger investment in new manufacturing equipment and infrastructure. As recently as 2014, one could count 7 companies capable of processing at the leading edge of semiconductor manufacturing technology, which at the time was 20nm. But just last year at 7nm, or today at 5nm, only TSMC and Samsung are in high-volume production, with Intel close behind.
Speaking of Intel, the Intel Accelerated event in July provided updated definitions and targets for their technology roadmap, guiding the technology node INTEL 7 to have products later in 2021 and forecasting themselves to be in a position of industry leadership by 2025 on the INTEL 18A node. According to Intel, the INTEL 18A node (“A” for angstrom) would require the industry’s first implementation of ASML’s most-advanced high-NA EUV photolithography.
A natural consequence of high-tech competition is that leading-edge foundries rarely reach the same technology position simultaneously, creating further asymmetry in the market. The first foundry to reach 5nm, for example, caught a great deal of attention from customers seeking to bring their designs to the latest node and capture improvements in speed and power consumption. High-profile customers are willing to pay a premium for this early volume on a new foundry node, while other customers must wait for the technology to become more mature and/or less scarce in order to sign on.
The ongoing trade dispute between the United States and China has created conditions that incentivize atypical business operations. One of these behaviors is the stockpiling of components. Parties directly restricted from access, like Huawei and Hisilicon, would have sought to increase their component inventories ahead of September 2020 when they lost access to TSMC output. Other OEMs not specifically restricted would have been wise to follow a similar strategy, though to a lesser degree. In their cases, using inventory as a buffer against potential supply chain risk outweighs the financial benefits of running with a leaner inventory. As the global trade posturing continues, so will this behavior.
COVID-19 and the global response continues to leave its mark as well. Initially, manufacturing lines and supply chains were immediately disrupted by lockdown enforcement. As local governments and companies adapted to navigate operations within a pandemic, there were new uncertainties in demand as a looming recession threatened to squelch any appetite for spending. A clear example is in the automotive space, where OEMs anticipated that economic uncertainty would bring a drop in demand for new vehicles and slowed down their manufacturing lines and reduced component orders. However, there was in fact a boost in demand for new automobiles as the pandemic drove down usage of public transportation and ownership of a personal vehicle promised to be an oasis from the locked-down world. By the time auto manufacturers had responded to this, foundry capacity had been allocated elsewhere.
We can also attribute a boost in consumer device demand back to COVID-19’s effects on the global economies. As schools and offices adapted policies to enact remote learning and working, students and employees had to reassess the capabilities of their home PCs, laptops, and tablets. Any home systems that were substandard or dated, were prime candidates for upgrade. In many learn-from-home settings, brand new systems (like Chromebooks) were sought to meet the minimum requirements for students to navigate the new remote-learning environment.
In addition to the effects of the pandemic, we saw exacerbating circumstances place further strain on the technology industry. Over the last year, Taiwan has received only a fraction of its typical rain accumulation, placing the largest hub of semiconductor production into a state of water insecurity. Significant winter storms in Texas, USA and a fire at Renesas’ Naka factory in Japan added bumps in the road for an industry trying to accelerate out of a COVID-induced stall.
Taking all these factors together paints the picture of an industry running too lean on existing capacity. Thus, enter TSMC, Intel, and Samsung (and others) seeking to boost their own capacity to exceptional degrees. Even recently, Micron Technology announced the sale of its Utah fab for $1.5 billion to Texas Instruments, which constitutes one of the quicker ways to add capacity: retooling an existing facility for a new type of devices.
Because the end-systems and supply chains of today are so highly integrated, the semiconductor shortage also impacts the shipment of devices which do not have a scarcity of capacity. For example, if an automobile OEM is unable to assemble a certain model due to the lack of a key component, then there are dozens, if not hundreds, of other components that the OEM does not need to procure. This dynamic plays out in a number of different ways: Microcontrollers, or MCUs, have not suffered as much from the competition for cutting edge technology nodes, but the MCU market has been affected by disasters and the supply chain impacts on the reduction in end-systems reducing demand, especially in Automotive which had evolved into a ‘just in time’ supply chain. Despite this, we estimate the MCU market to ship $17.5 billion worth of devices in 2021, growing to $24.2 billion by 2026.
Similarly, the investments made by foundries toward more diversified wafer production capacity will enable further scaling of transistor sizing and continued growth in the processor (CPU + APU) markets to over $140 billion by 2026, we estimate. The front-end foundry services we expect for these markets should exceed $34 billion, plus another $13 billion of value from Independent Device Manufacturers (IDMs). Adding GPU and FPGA to the mix and total front-end processing revenue should reach $60 billion in 2026.
While these challenges represent growing pains for ever-complex technology industry, the silver lining here is that these fragilities have been exposed and are now receiving attention, which will make for a more robust technology industry going forward. In theory, manufacturing lines and supply chains should be more hardened against future strain with new policies and investments, including those from TSMC, Samsung, and Intel.
Of course, investments made today will take some time to come online and may only be fully available for manufacturing starting in 2023. No one can say exactly what the state of the semiconductor industry will be like two or more years from now, and there is a risk that the industry would not need the additional capacity by then. However, if the trends hold in key technologies and end systems, such as automobile electrification, AR/VR, and micro-LED for example, then wafer demand will be substantially higher, and these investments will have been prudent. The ever-dynamic semiconductor industry is never dull!
About the authors
John Lorenz is a Technology and Market Analyst, Computing & Software within the Semiconductor, Memory & Computing division at Yole Développement (Yole), part of Yole Group of Companies. John is engaged in the development of market and technology monitors for the logic segment of advanced semiconductors, with an initial focus on processors. Prior to joining Yole, John held various technical and strategic roles at Micron Technology.
On the engineering side, his roles included thin film process development and manufacturing integration on DRAM, NAND, and emerging memory technologies and industrial engineering / factory physics for the R&D fab.
On the strategic side, John ran the memory industry supply & capex model for corporate strategy / market intelligence and established the industry front-end costing model within strategic finance.
John has a Bachelor of Science degree in Mechanical Engineering from the University of Illinois Urbana-Champaign (USA), with a focus on MEMS devices.
Tom Hackenberg is a Principal Analyst for Computing and Software in the Semiconductor, Memory and Computing Division at Yole Développement (Yole). Tom is engaged in developing processor market monitors and research into related technology trends. He is currently focused on low and ultralow power solutions such as MCUs. Tom is an industry leading expert with more than a decade’s experience reporting on markets for semiconductor processors including CPUs, GPUs, MPUs, MCUs, SoC ASICs & ASSPs, FPGAs and configurable processors. Tom is also well-versed in related technology trends including IoT, heterogeneous processing, chiplets, AI and edge computing.
Prior to joining Yole, Tom was a principal analyst at OMDIA, IHS Markit and began processor market research in 2006 for IMS Research. He worked with market-leading processor suppliers developing both syndicated and custom research. Tom holds a BSECE from the University of Texas at Austin specializing in Processors and FPGAs.
Related reports and monitors
For the first time, the processor monitor is including FPGA, CPU, GPU, and APU including all the IDMs, Fabless companies, and Foundries in the business.
The global MCU market will reach $24.2 B in 2026.
Penetrating everyday products will see the market for AI technologies for the consumer market reach $5.6B in 2026.
Popular data-intensive applications and the spread of AI applications are boosting server computing market to reach US$33B in 2025.
5G mobile processor using Samsung’s 5nm FinFET process with embedded packaging.
Intel’s 11th-generation core processor using its latest 10nm SuperFin process.
Ambarella vision processor using Samsung’s 10nm process targeting video surveillance, home security camera and machine vision applications.
Related Reports and Monitors
AMD Ryzen 5 Pro SoC Central Processing Unit
Reverse Costing - Structural, Process & Cost Report
Epitaxy Equipment for More than Moore 2021
Market & Technology