Report suggests DOJ is examining anticompetitive pricing strategies to lock up customers, lock out competitors.
Bloomberg News reports that the Department of Justice Antitrust Division has issued subpoenas to Nvidia Corp., the leading provider of artificial intelligence processors, suggesting an investigation into whether the company’s conduct in the burgeoning AI chip market is anticompetitive.
According to anonymous sources cited by Bloomberg reporters Ian King and Leah Nylen, DOJ has sent legally binding requests to Nvidia and other companies seeking information related to its business practices, a recent acquisition, and its dealings with customers and competitors. The reporters wrote of DOJ’s concerns that Nvidia uses its power to make it difficult for customers to switch to new suppliers and retaliates if a customer doesn’t use its AI chips exclusively. A publication called The Information first reported on the government’s probe after Nvidia’s competitors complained about its pricing strategies.
Nvidia has acknowledged that it is in contact with the DOJ but has not confirmed whether it has received subpoenas. In a statement released after the Bloomberg report, the company said it is committed to complying with all applicable laws and regulations. Emphasizing its focus on innovation and fair competition, the company told Fortune that it “wins [customers] on merit” and does not try to bar customers from buying chips from competitors.
Nvidia’s Place in the Market
Nvidia’s high-performance graphics processing units (GPUs) have become essential for training and running AI models, making them a coveted resource for technology companies and researchers.
In a June 2024 overview of the market, Statista put the AI chip market at $53.7 billion in 2023, anticipating growth to $71 billion this year. Nvidia, Statista wrote, “became a member of an exclusive club of tech companies valued at over $3 trillion, joining Microsoft and Apple in this feat in June 2024.” Nvidia’s chips are used to train and run various large language models, including the one developed by OpenAI’s wildly popular ChatGPT service, an endeavor that linked Nvidia GPUs with a Microsoft supercomputing data center.
In April 2024, Nvidia announced its purchase of Run:AI, a leading Israel-based provider of AI workload management and orchestration software. This acquisition strengthened Nvidia’s full-stack AI platform strategy and could potentially be subject to scrutiny as part of the DOJ’s investigation, Bloomberg reports. Consummation of the estimated $700 million deal is now expected to be delayed.
Nvidia has been in acquisition mode for several years. In February 2023, it acquired artificial intelligence startup OmniML, which focuses on miniaturizing machine learning applications – including large language models – so they can run on “edge devices,” which normally are routers that provide authenticated access to faster “backbone and core” networks.
Nvidia’s attempt to acquire Arm Ltd. in 2020 – owned by SoftBank of Japan – was ultimately abandoned in the face of regulatory challenges, but it demonstrated Nvidia’s ambitions to expand beyond AI chips and into the broader semiconductor market. See our previous post.
Nvidia also sees educating tech professionals as a path to growth. An acquisition helped it launch Nvidia Deep Learning Institute, a training and education platform for developers and researchers. The company boasts capacity to train hundreds of thousands of professionals.
Nvidia’s primary competitors in the AI chip market include:
- Advanced Micro Devices (AMD) is a direct competitor to Nvidia, offering its own line of GPUs and CPUs for AI applications, including its Ryzen processors and its Vitis software for developers.
- While Intel is primarily known for its CPUs, it has also invested heavily in AI chips, including its Habana Labs acquisition.
- Google has developed its own AI chips, known as Tensor Processing Units (TPUs), which are used in its data centers for training and running AI models. Amazon and Microsoft are also aggressively innovating in this space.
- Qualcomm, primarily known for its mobile chipsets, has also been expanding into the AI chip market with its Snapdragon AI processors.
- Worth noting is another SoftBank company. As a UK startup, Graphcore developed its own AI processor architecture, targeting high-performance computing applications. It was acquired by SoftBank Group Corp. in July and operates as a wholly owned subsidiary. According to one source, the Wall Street Journal reported that SoftBank paid more than $600 million, “well below the company’s peak valuation of $2.8 billion from a 2020 funding round.”
Related posts from MoginRubin:
Google’s A.I. Service Draws Early Copyright and Privacy Class Action