中文 English

End In Sight For Chip Shortages?

Some segments are normalizing, others may be impacted through 2022.

popularity

The current wave of semiconductor and IC packaging shortages is expected to extend well into 2022, but there are also signs that supply may finally catch up with demand.

The same is true for manufacturing capacity, materials and equipment in both the semiconductor and packaging sectors. Nonetheless, after a period of shortages in all segments, the current school of thought is that chip supply may return to relative normalcy by mid-2022, despite some product shortages like automotive chips, which could persist throughout 2022. This depends on several economic factors, however, so all of this could change overnight.

It’s been a chaotic period in the semiconductor industry. In early 2020, the business looked bright, but the market dropped following the Covid-19 outbreak. Throughout 2020, countries implemented various measures to mitigate the outbreak, such as stay-at-home orders and business closures. Economic turmoil soon followed.

By mid-2020, the IC market bounced back, as the stay-at-home economy drove demand for computers, TVs, and other consumer electronics. Shortages for consumer chips and select IC packages surfaced. Then, in the first half of 2021, demand for cars, smartphones and other products spiked, causing chip shortages in those sectors. Today, many chip types are in tight supply with long lead times, while a few others are easier to find. It depends on the chip and vendor.

Worldwide chip manufacturing capacity is also tight, especially for more mature processes in 200mm fabs. For some time, 200mm fab capacity has been sold out, and the situation isn’t expected to change anytime soon. And now, many foundry customers are bracing for another round of price hikes across the board. Meanwhile, on the packaging front, some package types will remain in short supply with tight capacity in many areas. Lead times for select equipment are long.

It’s not all doom and gloom for the shortage situation for semiconductors. “The supply and demand situation is expected to be mainly addressed in H1/2022 or H2/2022, with the exception of some products,” said Handel Jones, CEO of IBS, in a new report. “A number of factors have contributed to the strong demand for semiconductors. However, some factors that contributed to the past growth in demand are weakening because of market saturation. The buying power of consumers will weaken due to the reduction of stimulus and impact of high inflation.”

There are also some ominous signs for suppliers. Over-capacity may occur sometime in the second half of 2022 or 2023, depending on the product, according to IBS.

It’s impossible to explain the dynamics of every semiconductor product or package. Each one has its own supply/demand scenario. But there are several key products that present some insights into the situation. Those include application processors, microcontrollers (MCUs), power management ICs (PMICs) and WiFi chips as well as various packaging technologies.

Fab landscape
Over the years, the IC industry has seen its share of upturns and downturns. The current upturn is one of the biggest in recent memory. In total, the semiconductor market is expected to reach $542.55 billion in 2021, up 21.62% over 2020, according to IBS. The market is projected to grow 7.13% in 2022, IBS predicts.

The wafer fab equipment (WFE) market is expected to grow by 40% in 2021, according to TEL. “Significant expansion in the WFE market is expected on a sharp rise in demand for leading-edge logic and memory,” said Toshiki Kawai, president and CEO of TEL, in a presentation.

Nonetheless, the semiconductor industry designs and manufactures a multitude of different chips, such as analog, GPUs, MCUs, memory, microprocessors and power semiconductors. GPUs, processors and other advanced logic chips are produced in 300mm fabs using various process technologies ranging from the 16nm/14nm to the 5nm nodes. (A process technology is the recipe used to manufacture a given chip in a fab. A node refers to a specific process and its design rules.)

From 16nm/14nm to 5nm, chipmakers rely on finFETs. “As compared to prior planar transistors, the fin, contacted on three sides by the gate, provides much better control of the channel formed within the fin,” said Nerissa Draeger, director of university engagements at Lam Research.

300mm fabs also are used for mature process nodes ranging from 65nm to 28nm. Meanwhile, other chips are manufactured in older 200mm fabs using processes ranging from 350nm to 90nm. Many chips also are produced in fabs at even smaller wafer sizes, such as 150mm, 100mm and others.

At present, mature process nodes in both 200mm and 300mm fabs are tight, if not sold out. “In the last few years, there has been a surge in demand for a wide variety of chips that are made on 200mm and mature CMOS technology nodes ≥28nm whether that be on a conventional CMOS, bipolar CMOS DMOS or RF-SOI based process platforms. These devices include MCUs, PMICs, digital display driver ICs (DDICs), RF ICs, and the image signal processing (ISP) wafers needed to fabricate backside illuminated CMOS image sensors. This demand is also underpinned by technical trends in several market segments,” said David Haynes, managing director of strategic marketing at Lam Research.

“The supply issues for automotive ICs are well documented, but at the same time there is increased demand from consumer products, new 5G enabled devices and display applications,” Haynes said. “The situation is further compounded as many IDMs and foundries that make these chips make not one but several of these products. Historically, they have been able to rebalance fab capacity to address increased demand for a certain product type, but when demand for so many products are surging at the same time, it’s difficult or impossible to flex output in this way. Although there have been increases in global capacity for some device types, for example display drivers, recent reports suggest that the overall industry is yet to attain supply-demand parity.”

All told, foundry capacity is tight. “Looking into the fourth quarter, we anticipate wafer shipments and the ASP trend will remain firm. Capacity utilization across 8- and 12-inch facilities will continue to remain fully loaded,” said Jason Wang, UMC’s co-president.

For both trailing- and leading-edge nodes, fab capacity is expected to be tight for the foreseeable future. This depends on the process and vendor. “While we do not rule out the possibility of an inventory correction, we expect TSMC’s capacity to remain very tight in 2021 and throughout 2022,” said C.C. Wei, CEO of TSMC, in a recent conference call.

Summarizing the situation, Samuel Wang, an analyst at Gartner said: “Foundries are mostly booked for 1H 2022. Some have long-term agreements signed with fabless customers for 3-4 years. Gartner’s assumption is that the normalcy of chip inventory will be reached in 2Q 2022. Shortages by smaller vendors on various components here and there could last longer.”

Application processor issues
Wireless, meanwhile, is the largest semiconductor segment, representing 40% of the business, according to IBS. In wireless, 5G smartphones and the associated infrastructure are the big drivers for many chips. In total, 5G smartphone shipments are projected to reach 578 million units in 2021, up from 225 million units in 2020, says IBS. Still, while 5G is growing in many regions, China’s smartphone market is slowing.

5G smartphones consist of a multitude of chips, such as the application processor, CMOS image sensors, memory, PMICs and RF. The application processor is a leading-edge device that incorporates the CPU, graphics and AI functions on the same chip.
Fig. 1: Application processor complexity continues to grow. Source: TEL/Wikipedia
Fig. 1: Application processor complexity continues to grow. Source: TEL/Wikipedia

Apple’s new iPhone 13 incorporates the A15 application processor, a 15 billion transistor design based on TSMC’s 5nm process. Many other phones incorporate Qualcomm’s Snapdragon 888, a 5nm system-on-a-chip.

These chips are produced by foundry vendors. Today, TSMC and Samsung are the only foundry vendors capable of manufacturing 7nm and 5nm chips, and both have 3nm in R&D. Intel, which recently re-entered into the foundry business, is ramping up 10nm and 7nm, with 4nm in R&D.

For some time, demand has been robust for leading-edge 5G-based application processors and chipsets. But there appears to be a slight shortfall in foundry capacity for these chips. “Shortages in wafer capacity will likely continue until Q4/2021 or Q1/2022,” IBS’ Jones said.

Just how long the capacity shortages will last depends upon several factors. “The technology for latest designs such as the Apple A15 and Qualcomm’s Snapdragon 888 is in 5nm, with a planned migration to 3nm in 2022,” Jones said. “If smartphone chipsets at 3nm ramp up in H2/2022, there is potential for some overcapacity in 5nm and 7nm in Q3/2022 or Q4/2022.”

That could change. Slated for 2022, Apple’s iPhone 14 was supposed to use TSMC’s 3nm process for the applications processor, analysts said. Now, however, the iPhone 14 is expected use 4nm. Apple’s iPhone 15, slated for 2023, will incorporate a 3nm application processor, analysts said. In other words, TSMC’s revenue ramp for 3nm is delayed until 2023.

All told, the 3nm production ramp for all parties is a moving target. “There are indications of a delay in ramping up 3nm wafer volume at both TSMC and Samsung,” Jones said.

That’s not the only issue for Apple and other smartphone vendors. In its most recent quarter, Apple suffered a $6 billion shortfall in sales due to chip and manufacturing capacity shortages. The problem didn’t involve access to leading-edge nodes, but rather a shortfall in chips at mature processes.

Apple’s sales were impacted by shortages in several areas, including OLED touch screen controllers, according to KeyBanc. Used to control the display, touch screen controllers are manufactured using mature processes.

Other chips at mature nodes also are in tight supply, including Wi-Fi 6 chips. Wi-Fi and some RF chips are produced at 28nm, 22nm, and 16nm. Shortages for Wi-Fi and other RF chips likely will continue until Q2/2022 and potentially Q3/2022, according to IBS.

PMICs also have been in tight supply for smartphones and other products. Used to control the flow and direction of electrical power, PMICs are manufactured using processes ranging from 180nm to 40nm. Shortages of PMICs are expected to continue through Q2/2022 or Q3/2022, according to IBS.

MCU woes
When it comes to chip shortages, the automotive industry has been the hardest hit sector. “The cost of shortages in MCUs and other semiconductors for the automotive industry is potentially $5.0 billion to $10.0 billion in 2021. However, automotive companies have been able to partly offset their losses by focusing on higher-end vehicles, which give higher prices and higher profit than lower-end vehicles,” IBS’ Jones said.

A car incorporates some leading-edge chips, but the vast majority of devices are based on mature nodes. The average car incorporates about $330 worth of semiconductor content, while hybrid electric vehicles can contain up to 3,500 chips with a value of about $1,000, according to U.S. International Trade Commission.

In 2020, demand for cars fell amid the Covid-19 outbreak, causing automotive vendors to reduce their chip orders. But by early 2021, the automotive business rebounded. Faced with low chip inventories, automotive vendors attempted to order more chips from suppliers. But vendors found themselves last in the queue for parts as smartphone suppliers and others jumped ahead of the pack.

Unable to secure enough chips to meet demand, many automotive OEMs lowered their sales targets. Some even have temporarily halted select production lines.

The problem went from bad to worse when a fab at Renesas recently caught fire, causing a shutdown and a disruption in supply. This fab, which recently moved back into production, makes MCUs and other chips.

Today, chip shortages for automotive are still problematic. There are signs that supply is catching up, but the situation could last well into 2022.

NXP, one of the world’s largest automotive chip vendors, is a bellwether here. “For NXP, lead times for about 75% of our automotive products continue to be above 52 weeks,” said Kurt Sievers, president and CEO of NXP, in a conference call. “In summary, we think the automotive supply/demand equation will continue to be out of balance through 2022.”

Some 42% of NXP’s products are manufactured using in-house fabs. The other portion is manufactured by foundries, such as GlobalFoundries, TSMC and others.

In automotive, MCU shortages are especially problematic. Providing the processing functions in cars and other systems, MCUs are processed in fabs at various nodes, such as 180nm through 28nm. Some MCU makers have fabs. Many outsource parts to the foundries.

In either case, there still isn’t enough capacity to meet demand for the automotive industry. With an average wafer price of $1,800 (300mm-equivalent), wafer capacity for automotive MCUs is approximately 440,000 wafers per month (wpm) today, according to IBS. But the industry requires another 100,000 wpm to solve the MCU shortages in automotive, according to the firm.

In response, foundries and MCU makers with fabs are adding capacity. At some point, capacity will catch up with demand. “As a result, the supply shortage for MCUs should be over in H2/2022. There is also potential for overcapacity (even with the long design cycle time) in automotive in H2/2022 or 2023,” IBS’ Jones said.

Packaging landscape
As before, the dynamics in the IC packaging market reflect the supply/demand picture in the semiconductor business. For some time, the ongoing surge in demand for chips has caused shortages of select manufacturing capacity, various package types, key components, and equipment.

Take manufacturing capacity, for example. ASE’s factory utilization rates were higher than 80% in the fourth quarter of 2020. By the second quarter of 2021, ASE’s utilization rates were 85% for packaging/assembly and close to 80% for test.

In 2021, the third quarter was also tight. “We’re still running at full capacity above 85%. And that will continue into Q4. And test is above 80%, and (that is) also continuing into Q4,” said Joseph Tung, CFO at ASE, in a recent conference call.

So packaging houses are expanding their capacities to meet demand. To meet demand, JCET has announced the official opening of the second phase of its IC packaging and testing facility in Suqian within China. “In the second half of this year, JCET’s global manufacturing centers continue to optimize their mass production technologies and operational efficiency,” said Li Zheng, CEO of JCET.

But a package consists of a number of components, which themselves are in short supply. For example, many packages consist of a base material or substrate.

“Substrate lead times have increased from 2-4 weeks to as long as 16-20 weeks,” said Rosie Medina, vice president of sales and marketing at Promex, the parent company of QP Technologies.

Today, there are about 1,000 package types in the market. Each one is targeted for a different application with its own supply/demand situation.

One way to segment the packaging market is by interconnect type, which includes wirebond, flip-chip, wafer-level packaging (WLP), and through-silicon vias (TSVs). Interconnects are used to connect one die to another in packages. TSVs have the highest I/O counts, followed by WLP, flip-chip, and wirebond.

Some 75% to 80% of today’s packages are based on wire bonding, according to TechSearch. Wire bonding mainly has been used for low-cost legacy packages, midrange packages, and memory die stacking. For years, packaging houses have added wire bond capacity, but recently there has been underinvestment in the arena. Then, in late 2020, demand for wire bond assembly skyrocketed, causing a shortfall of capacity. And the shortfall has extended throughout 2021.

“It’s been pretty tight for several quarters,” said Charles Shi, an analyst at Needham. “It goes back to the semiconductor shortage situation. What’s short is not so much the advanced process nodes. The shortages involve trailing-edge nodes, which tend to require wire bond. And so that leads to a big problem. In the wafer fab, the trailing-edge process nodes are short in terms of capacity. Then, on the other side of the coin, wirebonding capacity is also short. They are tied together as chips manufactured with trailing-edge processes typically get wire bonded.”

Wirebonders are used to make several package types, such as quad-flat no-leads (QFN). QFN belongs in the leadframe group of packages. A leadframe is a metal frame. In the production process, a die is attached to the frame, and leads are connected to the die using thin wires.

“Like everyone else, we are seeing shortages and extended lead times, specifically on the materials side, which impacts several technologies. For example, we are waiting much longer for grinding blades for our wafer backgrinders, and delivery times have tripled for leadframes,” Promex/QP’s Medina said. “We have added new wire bonders, which increased our capacity and our ability to perform heavy wire and ribbon bonding. We’ve also launched our custom substrate development business. We’ve been steadily adding head count and extending shifts to handle customer demand.”

Others are also staying ahead of the curve. “Wire bonding shortages are being driven mainly by supply chain constraints on chip and materials. So far, Amkor is managing this with minimum impact,” said Siva Mohandass, senior vice president of the Wirebond/Power & Automotive business unit at Amkor.

Still, it’s challenging to obtain wire bond equipment. Earlier this year, wire bond lead times were 10 to 12 months, analysts said. At present, the lead times are around 3 to 6 months, they noted.

Besides wirebond, flip-chip demand is also robust. Flip-chip is used to develop BGAs and other package types. In the flip-chip process, copper bumps or pillars are formed on top of a chip. The device is flipped and mounted on a separate die or board. The bumps land on copper pads, forming electrical connections.

“Flip-chip demand is strong. People added a lot of capacity,” Needham’s Shi said. “Flip-chip equipment lead times are not as bad as wirebonders. That tells me the flip-chip capacity is certainly tight, but not as tight as wirebonding.”

Advanced packaging tool demand
Meanwhile, fan-out packaging, one type of WLP, is gaining steam in smartphones, watches and other products. In one example of fan-out, a DRAM die is stacked on a logic chip.

TSVs are used in advanced 2.5D/3D packages, which are aimed for high-end systems. In 2.5D/3D, dies are stacked or placed side-by-side on top of an interposer, which incorporates TSVs. The TSVs provide an electrical connection from the dies to the board.
Fig. 2: Different options for high-performance compute packaging, interposer-based 2.5D vs. Fan-Out Chip on Substrate (FOCoS). Source: ASE

Fig. 2: Different options for high-performance compute packaging, interposer-based 2.5D vs. Fan-Out Chip on Substrate (FOCoS). Source: ASE

AMD, Intel and others have been developing new 3D-like packages using the chiplet model. For chiplets, a chipmaker may have a menu of modular dies in a library. Customers then can mix-and-match the chiplets and integrate them in an existing package type or new architectures.

“Therefore, the system can be optimized by using the best processor components with an optimum performance/cost process node,” said Xiao Liu, senior program manager from Brewer Science.

Using this approach, Intel is stacking dies and bonding them using fine-pitch copper bumps. This is done using a system called a thermocompression bonder (TCB).

Demand for TCB systems is picking up. “ASM Pacific shipped a total of 250 TCBs worldwide as of February 2021, the majority of which, we estimate, went to Intel,” Needham’s Shi said. Besi, K&S and others also sell TCB systems.

Meanwhile, AMD is implementing a newer technology called copper hybrid bonding, which uses copper-to-copper interconnects rather than bumps for finer-pitch packages with more I/Os than traditional packages.

AMD will use TSMC’s hybrid bonding technology. TSMC is installing hybrid bonding equipment from Besi within its new facility in Taiwan, according to Shi. “In about a year or two, the hybrid bonding equipment market is expected to take off,” Shi said. “Besi announced that the company is building a new factory in Malaysia dedicated to hybrid bonding systems. The designed capacity of the factory is ~12-15 systems per month, or ~150 system per year.”

Conclusion
The chip/package supply chain is complex, with a variety of dynamics. It takes a scorecard to keep track of everything.

All of this is keeping the procurement groups at various companies busy for now, as well as for the foreseeable future.

Related stories
Chip Shortages Grow For Mature Nodes
Impact felt across many industries, including appliances, smart phones, cars, and industrial equipment.

Shortages, Challenges Engulf Packaging Supply Chain
Innovative business models emerge, but so does possibility of consolidation.

Automotive IC Shortage Drags On
Long lead times expected at least through the end of this year, as chipmakers scramble for solutions.



Leave a Reply


(Note: This name will be displayed publicly)