By the end of 2027, China could nearly double its available computing power for artificial intelligence — from the current 4.6 million to 8.8 million H100 equivalents. This forecast is contained in an analytical report by the Institute for Progress (IFP), which sheds light on the real scale of China's AI infrastructure.

The key paradox is that this growth, according to analysts' estimates, will be driven to a significant extent by American chips — despite the existing US export restrictions. Chinese companies are finding workarounds for supplies, stockpiling equipment, and actively adapting available accelerators to their needs.

What lies behind the numbers

The H100 equivalent is a convenient metric for comparing a heterogeneous fleet of equipment: from Nvidia's flagship GPUs to less powerful solutions and China's own domestic developments. Doubling this figure means that China is not merely offsetting sanctions pressure but is also forming one of the world's largest pools of computing for training and inference of AI models.

Such dynamics call into question the effectiveness of export controls as a containment tool. Restrictions slow access to cutting-edge chips, but they do not block the buildup of aggregate capacity — the market adapts through parallel imports, gray re-exports, and accelerated development of local analogues.

For the industry, this is a signal that the AI computing race is taking on a distinctly bipolar character: two large technological contours are forming, each striving for self-sufficiency. At the same time, Chinese giants retain access to a significant part of the American ecosystem, which blurs the boundaries of technological confrontation.

Cryptalist analysis: The forecast of a doubling of China's AI capacity points to the structural resilience of demand for computing even under sanctions pressure. For the market, this means continued high demand for accelerators and related infrastructure, as well as intensified competition for computing resources as a key pricing factor in the AI equipment segment.