Apple vs Huawei

Both Apple and Huawei are market leaders in high-end smartphones and are increasingly direct competitors. Moreover, both companies have developed their own AI chips in-house, with Huawei claiming that in some areas its chip is actually superior to that of Apple.

To illustrate the AI chip supply chain, we focus on Apple’s A13 Bionic chip (inside the iPhone 11 Pro) and Huawei’s Kirin 990 5G chip (inside the Mate 30 Pro). By detailing the input suppliers for these two AI chips, we show both a high degree of overlap in their supply chains and Huawei’s recent moves toward diversification.

Based on the breakdown of each company’s inputs, from design and fabrication to assembly and packaging, it appears that Huawei is more exposed to foreign suppliers in each segment of its AI chip supply chain. This dependence probably explains why Huawei is moving rapidly to secure domestic inputs, a drive that was further catalyzed by Washington’s latest actions to restrict US chip exports to the company.

Although Apple is less exposed to supply chain disruptions, neither is it entirely immune. That’s because, just like Huawei, Apple also depends on TSMC to manufacture its A13 chip. All of TSMC’s facilities capable of manufacturing these latest generation 7nm AI chips are located in Taiwan, which continues to be a geopolitical flashpoint between the United States and China. An overreliance on a single Taiwanese chip fabrication company carries supply chain risks not only for Apple and Huawei but also for the broader semiconductor industry as well.

Click on each of the chip components to see suppliers for Apple and Huawei.

AI_CHIP_case_study NPU NeuralProcessingUnit GPU GraphicsProcessingUnit CPU CentralProcessingUnit

Note: the countries denote where the companies are headquartered, not the location of production.

CHIP_ASSEMBLY_TEXT OSAT Manufacturing
Apple
USA
  • Apple developed its own NPU, known as a “neural engine.”
  • Since NPUs are usually designed for the AI algorithms needed for certain tasks, these chips are highly customized. Companies may follow Apple’s lead in bringing the design of these specialized chips in-house.
  • Apple has focused exclusively on AI chips for mobile devices rather than for cloud computing. One reason seems to be that Apple’s emphasis on user privacy means it does not want user data to be stored on the cloud but rather right on the device.
Huawei (HiSilicon)
China
  • Before 2017, Huawei relied on Cambricon, a China-based design firm, for AI chip design. But Huawei has by now already developed a unified neural processor architecture called “Da Vinci,” for both edge and cloud applications. And its Kirin 990 is the first chip to adopt this architecture.
  • Huawei also wants to establish itself in the AI platform ecosystem. It plans to launch MindSpore in 2020 to compete directly with Google’s TensorFlow.
Apple
USA
  • Apple previously used GPUs designed by UK-based Imagination, but has recently shifted to designing GPUs in-house, a model applied to all its products because it prefers end-to-end control.
  • GPUs are currently dominated by Nvidia and are crucial for AI applications such as facial recognition.
ARM & HiSilicon
UKChina
  • Same as the CPU, the chip’s GPU also uses ARM architecture but is designed by Huawei’s HiSilicon that was established in 2004.
ARM/Apple
UKUSA
  • Apple has collaborated with ARM since the 1980s, and its A13 CPU uses ARM architecture.
  • ARM’s dominance in mobile chips is mostly due to its core IP that balances efficiency with affordability, which is important because the AI chip represents 13% of the manufacturing cost of the iPhone 11 Pro Max.
  • To make the CPU better at handling ML tasks, the A13 chip has integrated an AI accelerator called the “neural engine” to process ML operations more efficiently.
ARM & HiSilicon
UK China
  • Huawei has taken a chapter from Apple and now relies on its in-house firm HiSilicon, established in 2004, for chip design (these designs are still based on ARM architecture).
  • The Kirin 990, which is 5G-capable, has eight CPU cores. These cores have different computing power, with the less powerful cores reserved for simpler tasks that consume less energy.
ASE (Advanced Semiconductor Engineering Incorporated)
Taiwan
  • ASE is the market leader in OSAT based on revenue.
  • Since OSAT, or the “back-end” of chip manufacturing, is considered relatively low-end, competition is much fiercer when compared to other segments of the semiconductor industry. ASE is now facing increasing competition from Chinese OSAT firms, including JCET.
JCET Group (Jiangsu Changjiang Electronics Technology Company)
China
  • Ranked third globally in 2018 based on revenue, JCET is the biggest OSAT firm in China and has also been an Apple supplier since 2015.
  • With significant market share in China, JCET is expected to benefit from the growth of the domestic semiconductor industry. It likely wants to follow the path of Taiwan’s ASE, which became the world’s largest OSAT firm thanks mainly to TSMC.
TSMC (Taiwan Semiconductor Manufacturing Company)
Taiwan
  • High start-up costs and highly specialized techniques mean that chip manufacturing is concentrated among a few players. Currently, only IDMs such as Samsung and Intel are in the same league as TSMC, though both still lag the Taiwanese firm in 7nm chip fabrication.
  • TSMC has six state-of-the-art GIGAFAB facilities, and all of its 7nm production facilities are currently located in Taiwan. The company has already begun working on next-generation 5nm technology.
  • Like most chip foundries, TSMC relies on ASML, a Dutch company, for one of the most expensive and crucal tools in wafer fabrication: the extreme ultraviolet (EUV) photolithography machine. TSMC is also a minority shareholder in ASML and collaborates with it on R&D.
  • In light of tightening US export controls, and to reduce its reliance on TSMC, Huawei has been looking to work with domestic foundries. China’s Semiconductor Manufacturing International Corporation (SMIC) has reportedly won Huawei orders at TSMC’s expense.