For AI tools users, discussions often focus on the latest version of Chatgpt or Gemini or Deepresearch. However, it is just part of the entire AI industry as well as the point of contact with users in the AI industry. at large. Leonardo Gambacorta and Vatsala SHREETI are placed in the “AI Supply Pan”. (March 154, 2025 International Settlement Thesis No. 154). The version of the AI supply chain is as follows.
Based on this structure, the core players of the entire AI ecosystem are:
Of course, this graph is an overview, not a detailed presentation, but often a useful way of gaining perspective. GAMBACORTA and SHREETI provide more details to the text. For example (see quotation and omitted graph):
hardware. Consider the most important hardware for AI applications, such as GPUs. NVIDIA, headquartered in Santa Clara, California, USA, provides services to most of the GPU market, and has a greater market share than 90%. The total margin is more than 70% and sales increased by 405% between 2023 and 2024. Initially, NVIDIA, which provided the service to the video game market, had a head start to utilize the GPU’s parallel computing capacity for the AI model. Over time, we have gained considerable intellectual property and significant reputation to strengthen our position as a market leader of the GPU. NVIDIA also generates complementary software in addition to the GPU itself. NVIDIA’s GPUs are provided in a monopoly bundle with CUDA, a parallel computing platform, so programmers and software developers can simplify and improve the GPU usage process. CUDA has become an industry standard for programmers and can only be used with NVIDIA’s GPU. …
Certainly, other companies, including new companies and large companies, are also activated in the AI hardware market. AMD (Advanced Micro Devices), Intel, and large technologies such as Microsoft, Google and Amazon are all producing AI microprocets to compete with NVIDIA’s GPUs for AI models and reasoning. Chinese companies such as Alibaba, Baidu and Huawei have begun to produce their own microprocessor, especially in light of designated scientific constraints. …
Cloud computing layer. Globally, the cloud computing market is dominated by Amazon Web Services (AWS) with a 31%market share, 24%Microsoft Azure and 11%Google Cloud platform. In the European Union (EU), the estimated market share of AWS and Azure in 2020 was more than 80%, and profit margins were high at 30%and 38%, respectively. In the case of IAAS segments, the most relevant of the AI model is the market much more concentrated. In 2023, the AWS, Microsoft Azure and Google Cloud platforms accounted for almost 74%of the global market. …
Education data. So far, the Frontier AI model has been trained using extensive public data. However, as the stocks of public data decrease sharply, the company is switching to other data sources. …
Basic model. At first glance, the basic model market is dynamic and with competitors. The market has over 300 basic models offered by 14 other companies. There is also a competitive business model. Some companies choose to provide a monopoly -based model (such as Openai and Google Deepmind), but other companies have adopted a relatively more open approach (especially open source LLAMA model and more deepSeek). Nevertheless, the Foundation Models market is currently dominated by a few companies such as Openai, Google Deepmind, Anthropic and Meta. Despite numerous competitive-based models in 2023, Openai’s GPT-4 accounted for 69%of the AI market in the global profit. Given the potential to realize the dynamic characteristics and efficiency of the market, the class can move quickly. …
AI application and layer for users. The layer faced by the user, the last stage of the AI supply chain, follows the playbook of the digital platform and mobile applications. After AI’s “CHATGPT MOMENT”, the applications built on the basic model have been spread in various economic sectors, including health, education, back -end processing and compliance, software development and other. Nevertheless, like the digital platform, there may be a risk of mechanics “all” winners “in the market for AI applications. Tracking the AI application market in all sectors is a herculin task, but the chatbot market can be beneficial. … [D]With a surge in similar interfaces, CHATGPT still explained 60% of the chatbot market in 2024, emphasizing the importance of the market for the first time.
One lesson of this overview is that if you believe that the AI tool will be an important driver who can move forward, a major economy like the United States should be interested in strengthening the entire ecosystem for the development and use of AI. The more difficult lesson is that using the technology that evolves very quickly, the most suitable approach to companies, users and policymakers is not clear. A company that appears to have a dominant market position in early 2025 may not have one in six months, one year or two years, regardless of whether the policymaker is taking action. As companies and strategies have increased, lessons of technology have been learned, and policy makers should be able to learn this lesson. In particular, policymakers are so busy that they are too busy to use this tool to spend time in the lobby, so they tend to listen to the greatest complaints instead of giving the beneficiaries equally.
