X
101871

A rattling of chains: How can chip suppliers avoid pandemics and politics?

May 4 2021
by John Abbott


Introduction


Why is there a chip shortage? Those familiar with the semiconductor sector will recall its cyclical nature – dips in the market have occurred roughly every five years since 1980, with the last one (at least the last in 'normal' times) in 2019. This time it feels different. We're now looking at a complicated mix of technological delays, geopolitics, natural disasters and supply chain complexities, issues that are unlikely to be resolved any time soon. Also, semiconductors are more central than ever to our everyday activities, and to the health of the global economy.

The chip shortage is doing much more than causing the chip companies themselves to negotiate some difficult financial quarters – it's also holding back key market segments like smartphones, games consoles, automotive, healthcare and manufacturing. The reasons include the sheer volumes involved, as in smartphones, and the increasing dependencies on multiple types of silicon in a product, such as cars. In 2016, the average car used between 25 and 50 processors. Today, it's in the hundreds, with electronics accounting for up to 40% of the total cost of a new car. The two big questions are: how can these new short-term and longer-term risks be mitigated, and what will the future of the semiconductor industry look like as these risks play out?

The 451 Take

The shift toward worldwide globalization is on hold for now, and it may not resume, at least in the form that we've grown used to. That's not good news for a technology world that for the past few decades has thrived on things like standardization, open source, and international scientific and technical collaboration. It's worth remembering that technology development is too complex to be carried out by a single country. The Pfizer vaccine illustrates this: it has 280 different components manufactured in 86 sites across 19 countries.

The immediate problem is that things can't be easily disentangled now, and highly distributed supply chains are at the center of this. Countries are now investing in the return of localized core silicon manufacturing capabilities. But local facilities can be equally disrupted by events. What's needed is more diversity in the supply chain, not less. Some short-term supply issues will soon be resolved, but longer-term factors that could end up reshaping the industry are already forming. Ongoing geopolitical shifts, health concerns, and climate change challenges lie ahead.

Context


The traditional boom and bust is caused by the contrast between the relatively short lifecycles of most semiconductor devices and the extended time it takes to bring new capacity on line. When a new technology or next major step-up of a key component is introduced, it generates strong demand, there's a boom in sales, and suppliers jump on the bandwagon.

But usually the supply is limited while the product matures, and this leads to tight availability, a rise in prices and high revenue for the chip vendors. After a while, supply becomes plentiful, with others jumping on the bandwagon, so inventories build, prices fall and so does revenue. Meanwhile, the product or technology that caused the boom in the first place heads toward obsolescence and something new emerges. Then the cycle starts again.

Coming out of the 2019 slump, it looked as if the rising demand for digital transformation – driven by the cloud, plus the emergence of new workloads requiring new architectures, such as AI, analytics and 5G – would drive the next boom. But other negatively influencing factors were already looming, starting with trade wars between the US and China. This was already problematic for advocates of globalization and free trade, but even more alarming because the real reasons for starting a trade war, and the ultimate goals, were unclear.

Then came the pandemic, causing factory closures, changes in working practices, and the unanticipated rapid acceleration of digital transformation projects, along with a series of natural disruptions and disasters that highlighted weaknesses in the globally distributed supply chains that hadn't previously been so apparent. Supply chains must be transparent, trustworthy and resilient to unexpected shocks to the system. They must also be efficient and flexible, and should incorporate lifecycle management and the circular economy, making more use of the previous generation of products and avoiding planned obsolescence.

Trade wars and geopolitics


The US trade deficit with China (amounting to $346bn in 2016) was the initial spur for the trade wars, and the plan to help rebalance the deficit cited such concerns as currency manipulation, intellectual property and forced technology transfer, export subsidiaries and limited market access, as well as poor labor and environmental standards.

But when it came to the IT sector, other reasons appeared to be the primary drivers. Telecom systems makers ZTE and Huawei were both designated national security threats, restricting the use of their products and technology in the US, and also stopping US companies from supplying them with key components and intellectual property.

With 5G in particular, it's hard to separate these security concerns from another likely motive – protection against competitive advantage. Huawei had forged ahead in its development of 5G infrastructure, and was making critical inroads into US and international telco operators by selling them infrastructure that, once established, would be very hard to replace.

The primary economic, national security and political concerns between the US and China (but also reflected globally) that affect the semiconductor sector are:

Economic concerns and trade. China's state-funded technology development is viewed by the US as a distortion of free market forces. China's outreach to supply IT to developing countries is seen as a means of building out a global presence by creating dependencies.

Trade restrictions. These have the potential to damage both sides. China is threatening to counter US technology controls with its own rare earth export curbs. Stockpiling by Chinese technology firms in advance of bans being enforced have contributed to the worldwide semiconductor shortage. US companies like Qualcomm have lost access to one of their main sources of older-design chips supplied to them by China's SMIC (Semiconductor Manufacturing International Corporation).

National security. Private Chinese companies have close links to the state, and the country's 2017 national intelligence law states that private organizations must support and cooperate in national intelligence work. Guaranteed continued access to critical materials or components can also be a national security issue.

Product integrity. Products might be modified as they move through the hardware or software supply chains for corporate or political espionage. The risks of letting a potentially hostile state supply the critical infrastructure of a rival country have to be considered.

Politics. The future status of Taiwan is unclear after recent events in Hong Kong. There has been an uptick in military activity from China. Taiwan is the dominant supplier of semiconductor manufacturing.

Competition. If trade restrictions are motivated by competitive positioning, things may turn out badly. Countries under threat from trade restrictions are now putting huge resources into their own domestic research and development, and may reach competitive parity with current technology leaders much more rapidly than they otherwise would have.

The pandemic and natural disasters


COVID-19 accelerated digital transformation globally. "We've seen two years' worth of digital transformation in two months," Microsoft CEO Satya Nadella said in April 2020. The shift to working from home placed a huge burden on IT, and security in particular, with many organizations turning to the cloud to take the strain. But for semiconductors, the main issues have been:

Rapid spikes in demand. Major increases in demand were fueled by sales of consumer devices (smartphones, games consoles, laptops) and the use of cloud services by people looking for home entertainment during lockdown.

Slump and recovery. There was business uncertainty in other sectors, leading to a massive slowdown, followed by a substantial rebound that occurred more rapidly than anticipated. Allocated chip orders were cancelled and couldn't be quickly reinstated due to high demand elsewhere.

Natural disasters. There have been a series of fires at chip factories in Japan (including the March 19 Renesas plant, a major supplier to the automotive sector), a severe drought in Taiwan (chip fabs require huge amounts of water) and severe cold weather in Texas.

Logistical blockages. These include a shortage of shipping containers, airfreight capacity limited by its use for vaccine shipments, the drop in passenger traffic, and the grounding of the Boeing fleets. There was also the Suez Canal blockage.

Supply chains and bottlenecks


The supply chains for semiconductors – spanning design, manufacturing, assembly and testing – are complex and distributed, and some believe they have become unbalanced. Figures from the Semiconductor Industry Association show that the US share of global chip production has declined from 37% in 1990 to 12% today, and it's continuing to head downwards. There are several reasons for this. Silicon Valley grew up as a center of manufacturing, but for decades now, the trend has been to move manufacturing offshore and focus on chip design rather than manufacture.

The original reason for that move was cost of production as the price of California land and labor rose. But Asia grasped the opportunity by being prepared to make huge investments on more narrowly focused activities. For instance, Taiwan Semiconductor Manufacturing Co (TSMC) was founded in 1987 as the world's first dedicated semiconductor foundry. By specializing in manufacturing, building up its volumes and optimizing its operations, it gradually saw competitors draw back because the required investment was too high for rivals to sustain.

TSMC's success led to the fabless semiconductor model and to the rise of companies such a NVIDIA that never owned their own fab. Apple, which now designs its own chips, is thought to be TSMC's largest customer, accounting for a quarter of its revenue. Fabless in now the predominant model. Of all the chip companies, only Intel has maintained and integrated design-manufacturing operations.

Equally important was the establishment of surrounding supporting ecosystems. This led to the original design manufacturing (ODM) model. It's the fabless model equivalent for systems manufacturers, which design their own laptops and servers and have them made by others such as Foxconn and Quanta.

A chain of companies, typically located close to each other, carry out tasks ranging from producing individual components through to subassembly, integration of chassis and motherboard, and further up the chain, to full assembly and testing. Apple has again benefited from this model for massive volume smartphone assembly, and it's something that would be very difficult to unravel at anything like a comparable cost.

With this background, the following supply chain danger points are currently top of mind for chip companies and the customers they supply.

Supplier concentration. TSMC and Samsung are the two largest and most advanced fab suppliers worldwide for CPUs and memory. Taiwan's political future is uncertain. Samsung is a divested organization that uses many of the chips it builds in its own products, competing with its customers.

Return to localization: Intel is investing $20bn in new US capacity. TSMC is also building new fabs in the US and elsewhere. US federal investments (through the Chips for America Act) will support this, and there are many equivalents elsewhere, including China and India.

Chipmaking equipment. Of all the potential single points of failure, this is perhaps the most critical. Only one company (ASML in the Netherlands) has the extreme ultra violet (UEV) machines capable of producing advanced process (10nm and beyond) CPUs. A handful of other companies, mostly in the US, make machines for older processes. The US government has restricted their sale to China, and their use by other companies selling on chips to China. Major fabs are experimenting with alternative production lines to avoid these dependencies.

Changing customers. Chip companies have seen their biggest customers transform into a small number of hyperscaler cloud providers and producers of consumer devices like smartphones. IoT, edge computing and automotive are rapidly growing areas, but more fragmented. From all these new markets, more flexibility and customization is a requirement.

Market access. Chip fabrication plants are massively expensive to build, and take up to three years to reach production status. Pure fab companies rely on high volumes to recoup the costs. If the markets are split along political lines and companies have to choose which ones to address, the economics may not always work. And it's not just about the location of a factory – skill levels, the surrounding ecosystem and licensing issues of associated technology, such as the software stack, also come into play.

Just-in-time manufacturing. Lean inventory practices that reduce the costs of components storage are the norm in sectors such as automotive. When these firms cancelled orders during the pandemic, the space capacity was snapped up by electronics companies, so was no longer available for car makers. Now that cars are essentially electronic devices, priorities and logistical dynamics have changed.

Social responsibility and eco-efficiency. Vying with all the other issues are factors such as social issues, like acceptable working practices in the factories, and the environment, such as the cost and environmental impact of moving goods around the world, often back and forth between the same countries, for multiple stages of production. Factory automation, supply chain trust and transparency, and new technologies such as blockchain have roles to play here.

This report can only scratch the surface of a complex and multifaceted subject, which also spans systems and software as well as semiconductors. A long-form report covering the issues in more depth is due to be published shortly.