“CATL Inside”? EV-Battery Maker Making a Name for Itself

Source: Daifuku. Data as of March 31, 2024.

In 2023, cleanroom-automation solutions grew to account for about a third of Daifuku’s sales—nearly on par with the intralogistics division that serves warehouses. But by 2025, cleanrooms will likely be the largest part of Daifuku’s business. The growth is being fueled by the changing production needs of its chipmaking customers, both in how and where they assemble their devices.

First, the how. This requires a brief tour through the physics of chipmaking as well as Moore’s Law, which is the observation that the number of transistors on an integrated circuit doubles every two years with a minimal rise in cost. As these devices get smaller, more can be added to the chips, giving them more processing power. However, innovations in transistor size have slowed in recent years, which is leading chipmakers to explore other solutions in a paradigm shift known as “More than Moore.” For instance, TSMC Chairman Mark Liu projects that within a decade, a graphics processing unit (GPU), a type of chip used for AI, will contain 1 trillion transistors, up from about 100 billion today, but the current pace of transistor miniaturization won’t allow such density. Unable to shrink transistors much further so that they fit side by side, chipmakers are instead stacking them in vertical layers, which saves space and improves processing efficiency by shortening the distance that signals must travel.

What does this mean for Daifuku? More of chipmakers’ budgets may be set to flow its way, as priorities shift from the front end of the production process, where the sizes of the chips and components are determined, to the back end, where transistor stacking and packaging takes place. In 2021, global semiconductor capital expenditures totaled nearly US$100 billion. About 80% of this went to the front end, as the focus on chip miniaturization over the past decade required foundries such as TSMC to make large outlays for expensive lithography machines that print the chips. Now, as stacking and packaging become more technical and important to increasing performance, greater investment will be needed to automate that stage of the process. By 2027, back-end solutions should account for 20% to 30% of Daifuku’s cleanroom sales. If back-end production becomes fully automated, it may be just as profitable for Daifuku as the front end, with operating margins of 10% and higher.

The other source of growth for Daifuku is a shift in where chip production takes place, as governments such as the US and Japan try to bring key tech manufacturing closer to home. The Semiconductor Industry Association (SIA) projects that wafer starts per month in the US will more than triple by 2032, taking the country’s share of global chip-fabrication capacity from 10% to 14%. Consequently, manufacturers will have to contend with higher labor costs and a shortage of skilled workers. This increases the need for automation, including at the back end as the focus shifts to packaging.

*Others include Malaysia, Singapore, India, and the rest of the world.

All values shown in 12-inch equivalents.

Source: Semiconductor Industry Association, BCG.

The US currently has only a small share of the US$95 billion assembly, test, and packaging market. But it wouldn’t make much sense to bring front-end chip-production work to the US only to have to ship components to Asia for the final steps. Therefore, the CHIPS and Science Act, passed in 2022, includes US$3 billion to build the country’s capabilities in advanced packaging. According to the SIA, companies such as Amkor Technology, Intel, Samsung, and SK Hynix are already planning packaging-related investments for facilities in Arizona, New Mexico, Indiana, and Texas.

These changes to the supply chain and chipmaking process, along with the broader advances being made in computing, are all favorable long-term trends for Daifuku.

The market for the driver-assistance technology is currently dominated by the Israeli company Mobileye. The company’s EyeQ system analyzes the input from a camera and sensors to alert drivers to dangerous traffic conditions and can initiate maneuvers such as evasive steering when a collision is imminent. As of last year, the company has captured about 70% of the private vehicle market with its basic driver assistance tools, with more than 50 automaker clients including Volkswagen, Ford, Toyota, and Chinese electric vehicle maker Geely.

Mobileye now hopes to leverage its success in basic driver assistance solutions to allow it to have similar success in more advanced levels of automated driving. Mobileye rolled out SuperVision, its Level 2+ driving assistance solution, in 2021. The system requires drivers to keep their eyes on the road, but they can let go of the steering wheel and let the automated system take over in specific circumstances such as on “all regular road types at up to 80 miles per hour,” according to the company.

In 2026, the company plans to release a true Level 3 autonomy system it calls Mobileye Chauffeur. That system would enable motorists to take their eyes off the road and their hands off the steering wheel in a wider array of traffic conditions, another step toward fully autonomous cars.

These more advanced systems come with extra risks for the company. If a car is involved in an accident while the automated systems are engaged, the manufacturer of the driver assistance tools can be held legally responsible for the incident. The added liability risk and the regulatory certification process, especially for Level 3 cars and above, are two of the reasons why few companies have ventured to bring advanced self-driving cars to market.

There is also strong competition in the market. The biggest threat is from large automakers that can afford to produce technologies for self-driving cars that could replace, or surpass, Mobileye’s offerings. An example is Mercedes-Benz, which in 2023 introduced the first Level 3 automation system approved for use in the United States. Chipmakers such as NVIDIA and Qualcomm also make competitive products.

One of Mobileye’s competitive advantages so far has been the low cost of its systems. Since they are mass produced and compatible with over 800 car models, the company has economies of scale that reduces the per unit cost of its products. The company’s basic driver’s assistance package sells to manufacturers for about US$50 per EyeQ unit, while the SuperVision product is sold for between US$1,000 and US$2,000 per system. Last year, most SuperVision units were sold to the electric vehicle maker Geely.

The cost of Mobileye’s automation systems for consumers depends on the manufacturer’s suggested retail price. For example, Ford recommends car dealerships charge consumers a US$600 to US$700 markup for the Ford Escape Active SUV, with Level 2 automation features.

This is compared with Tesla’s Level 2 offerings. While most Tesla car models are manufactured with the basic autopilot features pre-installed, consumers can upgrade to the enhanced Autopilot function for US$6,000 or the Full Self Driving package for US$12,000.

So far, orders for Mobileye’s advanced offerings have been below market expectations for 2024 as the company initially expected to ship 175,000 to 195,000 SuperVision units to a handful of large automakers including Porsche, China’s FAW group, Geely, and India’s Mahindra. But Mobileye’s stock edged up in late March after Volkswagen Group announced it would use SuperVision and Chauffeur in the future production of car brands such as Audi, Bentley, and Lamborghini, and also unveiled a partnership with Mobileye to rollout a robot taxi fleet in 2026.

Despite the positive update, investors would like to see more large Western automakers join the ranks of customers for the company’s new system. Mobileye says it’s in talks with 11 manufacturers who make 37% of global light vehicles. If the company wins more deals with automotive companies that essentially give up on their in-house autonomous vehicle systems, Mobileye could create a competitive moat to defend its long-term profits and market share.