The power map in artificial intelligence Who owns the companies, models, and chips?

The artificial intelligence sector is witnessing a global race that extends beyond competition between chatbots and generative models to encompass control of chips, data centers, cloud computing, and capital. At the heart of this race stand the United States and China, with a clear difference in the size of their companies, investments, and available computing power.

Hot Topics 

Musk proposes a framework for AI safety: "Let companies test each other's models

Need to know regarding how Kinsta can help your sites start making use of and become quicker for you?

Amazon Kindle Book : A Practical Guide for Married Couples

the processor industry, while Microsoft, Google, Amazon, Meta, and Oracle lead a large part of the infrastructure and cloud computing, along with model development labs such as OpenAI and Anthropic.

The importance of computing companies is clearly demonstrated in the data from Epoch AI, which estimated that Amazon, Google, Meta, Microsoft, and Oracle collectively owned about 71% of the total global AI computing power in the last quarter of 2025, compared to 63% in the first quarter of 2024. The source indicates that Google alone owned the equivalent of about 5 million H100 units of computing power, or about a quarter of the global total, according to the methodology of the study.

American power is not limited to computing power; partnerships link model developers with infrastructure providers. Microsoft and OpenAI confirm the continuation of their partnership, with Azure remaining the exclusive cloud provider for OpenAI's non-state interfaces, while Anthropic relies on Google Cloud, AWS, and Microsoft Azure.

Anthropic announced last April an agreement with Amazon to secure up to 5 gigawatts of computing power, with Amazon's investment in the company reaching $25 billion when accounting for new and potential investment.

As for market value, it changes with the movement of stocks daily. According to market data in September, Nvidia's value reached about $5.42 trillion, compared to about $4.15 trillion for Alphabet, $3.70 trillion for Microsoft, and about $1.98 trillion for Meta.

China: A local system expanding

On the other hand, China is building its own system of models and infrastructure, with companies like Alibaba, Tencent, Baidu and ByteDance, along with model development labs and companies like DeepSec.

Alibaba has a growing presence across its Qiwin model, cloud computing, and chips. This month, the company announced its GenWo V900 AI chip, stating that a single cluster built on it can support up to 500,000 cards for model training and operation. It also set a goal of exceeding 20 gigawatts of capacity for Alibaba Cloud's global data centers by 2032.