Nvidia is Building a Shield of Concentrated Power

Megan Kirkwood is a fellow at Tech Policy Press. This is the third in a series of three posts on Nvidia’s dominance in the AI industry.

This article is the final part of a three part analysis of Nvidia’s rise to dominance in the AI market. The first article in the series looked at the rise of the company from a gaming chip maker to industry giant. The second looked at how the company has verticalized across the tech stack, and spread horizontally across industries. It also considered how the company is developing partnerships with governments around the world. This final post looks at the antitrust scrutiny the company has recently attracted, and how those efforts, which have largely stalled, are now potentially jeopardized by the ambitions of those partner countries to build out AI infrastructure and get ahead in the so-called “AI race.”

Mapping antitrust efforts

In recent years, competition authorities have increased scrutiny of AI-focused companies, particularly the small number of incumbent firms with existing power in important digital markets that are now profoundly shaping the development of AI-related markets, such as Meta, Google and Microsoft. Increased regulatory scrutiny has also looked upstream in the AI tech stack. Competition authorities in the US, UK, EU, France, South Korea, and China have all launched inquiries into the business practices of Nvidia.

Last year, the United States Department of Justice delivered questionnaires to the company, escalating its scrutiny in September 2024 by sending Nvidia a subpoena seeking evidence as to whether the company had violated antitrust laws. Officials were apparently concerned that Nvidia makes it harder to switch to other suppliers and penalizes buyers who don’t exclusively use its GPUs. Regulators also questioned whether Nvidia gives preferential supply and pricing to customers who use its technology exclusively or buy its complete systems. Bloomberg reported that “Nvidia acknowledged allocating its chips to customers it deems most likely to use them quickly, prompting concerns that it has too much power over the market for semiconductors.”

Also last year, Nvidia also reported to the US Securities and Exchange Commission that the UK was examining its business practices. The UK’s Competition and Markets Authority (CMA) identified an “interconnected web” of over 90 partnerships and strategic investments involving the same firms: Google, Apple, Microsoft, Meta, Amazon, and Nvidia. The CMA’s main concerns were that:

• “firms controlling critical inputs for developing FMs [foundation models] may restrict access to shield themselves from competition,”
• “powerful incumbents could exploit their positions in consumer or business facing markets to distort choice in FM services and restrict competition in deployment,”
• and “partnerships involving key players could exacerbate existing positions of market power through the value chain.”

The CMA has not yet formally acted, but stated its intention to monitor “the competitive landscape in AI accelerator chips and the impact on the FM value chain as part of the next phase of work on our FMs initial review,” recommending that compute should be considered for investigation under the Digital Markets, Competition and Consumers Act.

In France, it was reported that antitrust enforcers raided Nvidia's offices on September 28, 2023, under suspicion the company was engaged in “anticompetitive practices in the graphics cards sector.” In a June 2024 press release, the Autorité de la Concurrence concluded its market study, finding that the company is likely abusing its dominance through “price fixing, production restrictions, unfair contractual conditions and discriminatory behavior.” The Autorité also expressed concern over the AI industry’s dependence on Nvidia’s CUDA programming software as well as Nvidia’s “investments in AI-focused cloud service providers such as CoreWeave are also raising concerns.”

Reuters then reported that the company was “set to be charged by the French antitrust regulator for allegedly anti-competitive practices … making it the first enforcer to act against the computer chip maker.” Later, the Autorité confirmed an investigation is underway.

The European Commission started to informally collect views on potentially anticompetitive practices in the sector for GPUs in 2023 to understand if there is a need for future intervention. Adding to the antitrust investigations, China’s antitrust authority had taken issue with commitments made by Nvidia when it acquired Mellanox, although some see the investigation as a retaliation against export bans made by the US. China has subsequently dropped its investigation following trade agreements with the US.

These antitrust investigations could potentially be jeopardized by politics, given that each of these countries and jurisdictions recognize the development of AI as a matter of geopolitical urgency. For instance, in the US, President Donald Trump announced his AI Action Plan, which is very clear in its intention to promote the nation’s AI champions to “accelerate American leadership.” Nvidia is integral to that mission. The EU has also announced significant investment into building AI factories in a push to build a ‘Eurostack’ of EU-owned and operated technology. This has not put off US tech giants, Nvidia included, from selling the EU “sovereign” AI infrastructure, despite very much still being US-owned and operated. Meanwhile, the UK wants to “mainline[…] AI into the veins” of the nation and has announced a slew of partnerships with various US tech firms to do so, including Nvidia.

The danger of concentrated power

If such nations wish to increase their infrastructure for computing, it will be very difficult to avoid Nvidia as a near monopolist in the market of AI-focused chips and related hardware, not to mention its software. Indeed, even efforts to build sovereign AI applications rely on Nvidia GPUs and sometimes its software tools. The sheer industry and research community capture may necessitate the use of CUDA and Nvidia GPUs. If nations choose to go “all in” on AI, will they concede investigations into Nvidia and the ways it has monopolized the market if they themselves become reliant on its infrastructure?

The lack of alternatives in the market gives a gatekeeper like Nvidia huge economic, informational and political power. As it reaps the benefits of AI infrastructure investments, Nvidia can shape the market through its allocation of vast resources, and can bend the geopolitical agenda to meet its needs. Indeed, Nvidia president and CEO Jensen Huang wasted no time taking his complaints to President Trump regarding GPU export controls issued under the previous administration, which have since been relaxed. In the latest deal, Trump has announced that Nvidia can sell its H200 chips to China in exchange for the US government getting a 25% cut of those sales. Meanwhile, Big Tech has been sure to advise countries like the UK to inform the implementation of the technology through public services and the wider economy.

This miniseries was intended to serve both as an explainer on Nvidia’s rise to power as well as its expansion vertically and horizontally, and thus to shine a light on the company’s growing strategic position in public and private infrastructure. Nvidia is able to influence what gets built and which technologies are prioritized, serving Huang’s vision of a fully automated world, the proliferation of “agentic AI” and very large language models. These applications all happen to be very computationally intensive and therefore require lots of Nvidia GPUs, benefiting the company immensely. However, a private company’s incentives may not reflect the goals of public institutions. AI Now Institute’s Kate Brennan, Amba Kak, and Dr. Sarah West write that injecting AI into public services will most likely mean increased automated decision-making, which “concentrates power among the deployers of the tech, leaving those on the receiving end more insecure, vulnerable, and unable to contest the determinations made by the ‘smart machine.’” Under the guise of efficiency, accountability can be lost.

Eryk Salvaggio writes for Tech Policy Press that not only is the increasing adoption of automated decision making eroding democratic accountability and decision-making procedures, “shifting government decisions to AI systems they must know are unsuitable, these tech elites avoid a political debate they would probably lose. Instead, they create a nationwide IT crisis that they alone can fix.” The more that government outsources to Big Tech, the more their technology is embedded in public infrastructure and the further their political power is solidified. This power means that dominant firms can set the agenda, protect their market position and establish an enduring shield against regulatory scrutiny and action.