Table Of Links Table Of Links Abstract Abstract Executive summary Executive summary Acknowledgments Acknowledgments 1. Introduction and Motivation 1. Introduction and Motivation 1. Introduction and Motivation Relevant literature
Limitations Relevant literature Limitations 2. Understanding the AI Supply Chain 2. Understanding the AI Supply Chain 2. Understanding the AI Supply Chain Background history
Inputs necessary for development of frontier AI models
Steps of the supply chain Background history Inputs necessary for development of frontier AI models Steps of the supply chain 3. Overview of the integration landscape 3. Overview of the integration landscape 3. Overview of the integration landscape Working definitions
Integration in the AI supply chain Working definitions Integration in the AI supply chain 4. Antitrust in the AI supply chain 4. Antitrust in the AI supply chain 4. Antitrust in the AI supply chain Lithography and semiconductors
Cloud and AI
Policy: sanctions, tensions, and subsidies Lithography and semiconductors Cloud and AI Policy: sanctions, tensions, and subsidies 5. Potential drivers 5. Potential drivers Potential drivers Synergies
Strategically harden competition
Governmental action or industry reaction
Other reasons Synergies Strategically harden competition Governmental action or industry reaction Other reasons 6. Closing remarks and open questions 6. Closing remarks and open questions 6. Closing remarks and open questions Selected Research Questions Selected Research Questions References References References Antitrust in the AI supply chain Antitrust in the AI supply chain There has been growing interest in more active antitrust measures in the technological sector. In this section, we highlight how this has been impacting relevant steps of the AI supply chain. We generally observe concerns of abuse of dominance, such as bundled sales and exclusive dealing, in the semiconductor industry and big tech companies. There have been some noteworthy merge controls in the semiconductor industry that we will describe, but none in the cloud providers or AI products directly. 4.1 Lithography and semiconductors 4.1 Lithography and semiconductors Until now, there has been limited antitrust litigation in the companies upstream of the AI supply chain. No acquisition by ASML and TSMC, respectively the most advanced lithography company and the most advanced foundry company in the world, has been challenged by US or EU authorities, and they were not adversely affected by any other kind of antitrust litigation in the past 25 years. In the upstream part of the industry, the most noteworthy antitrust case was from companies that supplied non-lithography semiconductor manufacturing equipment to chip fabricators. From 2013 to 2015, Applied Materials and Tokyo Electron attempted to merge for USD 29 billion, which raised concerns from authorities as this would be a horizontal integration of, respectively, the first and second-largest company in this market segment. The merger was eventually abandoned by the companies after the Department of Justice of the US raised competition concerns and rejected their proposed remedies (Department of Justice, 2015). The chip designing industry is receiving increasing attention from antitrust authorities. The FTC challenged Nvidia's acquisition of Arm Limited by USD 40 billion because of concerns that NVIDIA would gain excessive market power as it would have incentives to foreclose the licensing of Core IP owned by Arm to other chip designers. In 2022, NVIDIA terminated the proposed acquisition of Arm (FTC, 2022; TechCrunch, 2022). This acquisition was also under scrutiny in the European Union under similar concerts (Reuters, 2021). Additionally, the European Union is reported to have launched an early investigation into suspected anti-competitive abuses by Nvidia in the AI chip market (Tech Going, 2023) and, in France, Nvidia's offices were raided by the country's antitrust authority over suspicions of anticompetitive practices (Forbes, 2023). In the US in 2008, Nvidia settled a GPU antitrust class action lawsuit in 2008, which alleged a price-fixing conspiracy with ATI to fix, raise, maintain, and stabilize prices of graphics processing chips and cards (Bit-Tech, 2008). In 2005, Advanced Micro Devices (AMD) opened a private antitrust lawsuit against Intel around allegations of anticompetitive practices in the x86 microprocessor market. Filed in June 2005 in the United States by AMD, the lawsuit accused Intel of engaging in illegal practices to maintain a monopoly over the market, including offering rebates to companies for purchasing most of their microprocessors from Intel, and retaliatory actions against customers who engaged with AMD. The case culminated in a settlement in 2009, where Intel agreed to pay AMD $1.25 billion and adhere to a set of business practice provisions to enhance competition in the microprocessor market (AMD, 2009). Afterward, Intel also reached a settlement agreement with the FTC, which prohibited the company “from using threats, bundled prices, or other offers to exclude or hamper competition or otherwise unreasonably inhibit the sale of competitive CPUs or GPUs” (FTC, 2010). The legal battle was part of a broader global scrutiny of Intel's practices. South Korea and the European Union also investigating Intel's market behavior. In Japan, the Fair Trade Commission (JFTC, 2005) took action against Intel in 2005, accusing the company of offering rebates to five prominent PC makers—Fujitsu, Hitachi, NEC, Sony, and Toshiba—on the condition that they limit or cease purchases from Intel's competitors, primarily AMD (CNET, 2005). Following this, Intel agreed to a cease and desist order (NetworkWorld). Around the same time, South Korea's antitrust authority, the Korean Fair Trade Commission (KFTC), initiated an investigation into Intel's practices in 2005, culminating in a fine of Won 26bn ($25m) in 2008 for abusing its dominant market position in the country (Computer World, 2018). The European Commission, too, was probing Intel's market behavior, and in collaboration with Japanese authorities, was investigating possible antitrust violations. Another major block to merge in the semiconductor industry has been the block to the proposed merger of Broadcom and Qualcomm. The proposed $117 billion merger between Singapore-based Broadcom Ltd and U.S.-based Qualcomm Inc faced severe scrutiny from U.S. authorities, leading to its blockage by President Trump due to national security concerns, particularly fearing an erosion of U.S. mobile technology leadership to China's advantage (Reuters, 2018). Despite Broadcom's attempts to alleviate concerns by pledging to redomicile to the U.S. and not sell critical national security assets to foreign entities, the merger was halted, reflecting U.S. efforts to safeguard national and technological security in the semiconductor industry (PCMag, 2018). There has also been a major price-fixing scandal in the semiconductor industry. The DRAM cartel scandal emerged in the early 2000s with multiple major manufacturers of dynamic random-access memory (DRAM) being implicated. The US Department of Justice initiated a probe in 2002, responding to claims from US computer makers like Dell and Gateway regarding inflated DRAM pricing impacting their profits (Department of Justice, 2005). Samsung, Hynix, Infineon, Micron Technology, and Elpida pleaded guilty to their involvement in a cartel spanning 1998 to 2002. On a global scale, European antitrust regulators fined nine semiconductor manufacturers over €331 million in 2010, reflecting actions that took place in 2002 (European Commision, 2010). The scandal saw criminal fines totaling more than $730 million against the DRAM cartel members, marking at the time the second-largest total amount of fines ever imposed in a U.S. criminal antitrust investigation (EDN, 2007). 4.2 Cloud and AI 4.2 Cloud and AI While big tech companies have been in increased scrutiny by antitrust authorities, especially for anti-competitive behavior in price-setting and damaging its competitors in their platforms, there has been no noteworthy antitrust litigation directed impacting the development of frontier AI systems. DeepMind acquisition by Google, for instance, was approved without conditions, with no significant information publicly available of antitrust authorities raising concerns about this deal. 4.3 Policy: sanctions, tensions, and subsidies 4.3 Policy: sanctions, tensions, and subsidies From their inception, the semiconductor and AI sectors have often been viewed as strategic markets for government involvement. For example, Silicon Valley's growth was partly fostered by DARPA contracts (see, e.g., Heinrich, 2002). Recently, both the U.S. and EU governments have shown a renewed focus on subsidizing the chip industry. We overview these actions for completeness, but they are not the focus of this paper. 4.3.1 Sanctions and geopolitical tension 4.3.1 Sanctions and geopolitical tension The semiconductor industry has faced significant tension. In 2019, due to national security concerns, the US imposed sanctions on China’s Huawei and pressured other countries not to adopt their 5G technology. In 2023, the US imposed further restrictions on the Chinese technology industry, banning the export of advanced AI accelerators to China and blocking the use of US technologies. The US also pressured allied countries crucial to the semiconductor supply chain. In response, China launched initiatives to expand its autonomy in the semiconductor industry. The key open question is how long it will take China to develop EUV technology or other comparable technologies independently. 4.3.2 Subsidies and industrial policy 4.3.2 Subsidies and industrial policy Both the US and the EU passed Chip Acts, multi-billion dollar subsidy plans to foster the development of the semiconductor industry within their jurisdictions. The US Chip Act is a $280 billion plan to boost semiconductor research and production in the US. It offers $52 billion for chip manufacturing, $24 billion in tax credits for chip tools, and $200 billion for scientific research and innovation. The act also aims to enhance US national security.(Zimmerman, 2022) The EU Chip Act is a €43 billion ($47 billion) project to support Europe's semiconductor technology and uses. It aims to enhance chip innovation, understand global supply chains, and fill the talent gap in the field, while also targeting Europe's digital and environmental goals (European Commission, 2023). The Netherlands has ASML, the top maker of chip-making machines. For national security reasons, the Dutch government has limited its sales to some countries, including China (Associated Press, 2023). Japan's Rapidus, formed in 2022 with the backing of eight major Japanese firms (Denso, Kioxia, MUFG Bank, NEC, NTT, SoftBank, Sony, and Toyota), aims to make advanced 2-nanometer chips by 2027. Rapidus has a tech deal with IBM and received an extra $2 billion from the Japanese government (Kizuna, 2024). South Korea introduced the "K-Semiconductor Strategy" with a $280 billion investment to boost its semiconductor sector through R&D, subsidies, and tax benefits until 2030. It focuses on national security and enhancing 5G supply chains. The "Semiconductor Cluster" project centers around SK Hynix and Samsung campuses to advance chip technology, backed by government support (Kim et al., 2023). Taiwan leads in chip production, mainly due to TSMC, the top chip-making company. TSMC makes chips for major tech brands and has invested heavily in advanced chip processes. The Taiwanese government actively helps semiconductor companies secure land, water, and electricity (Chiang, 2023). However, they face political issues from China, which views Taiwan as its own. China's strategy for semiconductor dominance includes the "China 2025 Plan," targeting 70% self-reliance in semiconductors by 2025. The "Big Fund," initiated in 2014, offers $21 billion in state-supported funds to finance domestic chip endeavors and encourage the acquisition  of foreign expertise and technology. Additionally, a forthcoming $143 billion investment package aims to further enhance China's semiconductor capabilities, prioritizing the production and innovation of advanced chips (Allen, 2023). Author:
Tomás Aguirre Author: Author: Tomás Aguirre This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. available on arxiv n arxiv

Part of HackerNoon's growing list of open-source research papers, promoting free access to academic material.

Dell

Google

Intel

Antitrust Pressure Builds Across the AI Chip Supply Chain

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

15 Common Types of Unethical Behavior Found in Open-Source Projects

As AI Accelerates, Regulators Race to Understand a Rapidly Integrating Supply Chain

How Big Tech Built the Modern AI Supply Chain

How the AI Supply Chain Evolved From Transistors to Frontier Models

How Big Tech Is Locking In the Frontier AI Supply Chain

The Noonification: Charting a Path for Global AI Regulation (11/14/2023)

15 Common Types of Unethical Behavior Found in Open-Source Projects

As AI Accelerates, Regulators Race to Understand a Rapidly Integrating Supply Chain

How Big Tech Built the Modern AI Supply Chain

How the AI Supply Chain Evolved From Transistors to Frontier Models

How Big Tech Is Locking In the Frontier AI Supply Chain

The Noonification: Charting a Path for Global AI Regulation (11/14/2023)

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps