Chip War: NVIDIA, ARM, and Technology Concerns with Semiconductors
Our previous coverage of semiconductor shortages discussed the global value chain of the semiconductor industry and the compartmentalization of chip making that exposes the industry to the risk of large-scale shortages.[1] This blog post will explore how a rush into the high-margin chip market resulted in the current shortage crisis and made it harder for NVIDIA to acquire Arm.
The automotive industry has been the hardest hit industry of the global chip shortage.[2] Unfortunately, the industry forecast is that carmakers will continue to suffer in the upcoming years because semiconductor companies are reluctant to restructure their factories for automotive chip production, which is not as profitable as the production of central processing units (“CPUs”) and graphics professing units (“GPUs”) for computers and mobile devices.[3] In addition, because artificial intelligence and machine learning have significant value creation-potentials for semiconductor companies, chip investments have been concentrated on making higher-margin, smaller chips for these purposes[4]
NVIDIA, the largest GPU designer in the world, seeks to strengthen its position in the semiconductor industry by purchasing British chip designer Arm at $40 billion from SoftBank.[5] NVIDIA and Arm informed the media that their proposed merger would create “the world’s premier computing company for the age of AI.”[6] Yet several major tech companies are vehemently against the proposed transaction.[7] International governments also raise concerns: England opposes the deal, and China is reluctant.[8] So what is it about NVIDIA and Arm that makes this deal so challenging?
NVIDIA, the proposed buyer, primarily designs GPUs, which have been especially profitable and useful as many tech companies use NVIDIA GPUs to process big data in their data centers. GPU’s parallel processing architecture makes it uniquely suitable for processing big data and even mining cryptocurrency as the architecture handles thousands of tasks at once. In contrast to GPU’s parallel processing architecture, CPU’s serial processing architecture handles much less threads at once and is therefore less fit for big data processing.[9] For this reason, NVIDIA as the biggest GPU maker has become one of the hottest and most valuable semiconductor companies in recent years, even surpassing CPU giant Intel’s market cap.[10].
The proposed seller, Arm, is a British chip designer whose chip architectures dominate the mobile processor market.[11] Arm designs architectures and licenses them to companies like Qualcomm and Samsung.[12] Arm’s presence is also growing in the personal computer market as Apple abandoned Intel’s x86 and adopted Arm’s architectures to design its own M1 chip.[13] In addition, as many other tech giants like Google, Facebook, Huawei, Microsoft, and Tesla enter into the chip making business to strengthen their artificial intelligence capabilities, Arm’s architectures may become even more valuable.[14] In fact, there is a rumor that Google plans to use Arm’s architecture to design chips.[15]
Because Arm has powerful architectures that it can license to major tech companies, there is a huge antitrust concern over NVIDIA’s acquisition of Arm. Because both NVIDIA and Arm are semiconductor companies, the deal poses a global opposition. The U.K. government has a “concern over national security, the loss of crucial homegrown technology, and key roles to a foreign buyer.”[16] Qualcomm, a rival chipmaker, told the Federal Trade Commission, the U.K.’s Competition and Markets Authority, and China’s State Administration for Market Regulation that it opposes NVIDIA’s purchase of Arm.[17] Qualcomm is worried that NVIDIA will become a “gatekeeper” to Arm’s intellectual property that has been so fundamental to the semiconductor industry.[18] Qualcomm itself also has an Arm-based product, Centriq.[19] However, there is also a concern that if NVIDIA fails to get approval for the acquisition of Arm, Softbank—Arm’s current owner—will have difficulty finding another suitable buyer who can also spend $40 billion to purchase Arm. At the very least, technology companies that are breaking into the chip-making business will not be able to purchase Arm over the same antitrust concerns that NVIDIA currently faces. This perhaps leaves financial institutions as sole potential buyers—nonetheless, mobilizing $40 billion is no small deal for them either.
As more tech giants break into the high-margin chip making business and regulators closely observe the industry, the global semiconductor industry will continue to be very competitive. Global regulators and Qualcomm worry that NVIDIA’s acquisition of Arm will discourage competition—NVIDIA may restrict the granting of Arm’s architecture licenses to NVIDIA’s competitors or raise licensing fees while strengthening its market position through Arm’s highly sought-after technology. In contrast to the vibrant landscape of the high-margin chip market, the low-margin auto chip market is in despair as chip foundries have shifted their focus to producing more profitable chips. Nevertheless, it is fascinating to observe how a competition in the computer and mobile chip market has resulted in chip shortages in the auto industry that, in turn, affects so many companies and consumers.
Footnotes
