Morgan Stanley significantly lowered its forecast for Nvidia GB200 shipments in 2025 from the previously projected 30,000-35,000 to 20,000-25,000 units in its latest report, with shipments in the worst case scenario likely to be less than 20,000 units.

The cut may lead to market impact on GB200 supply chain of $30 billion to $35 billion, bringing great pressure to the relevant supply chain and semiconductor companies.

As for the reasons for the cut, Morgan Stanley said that Microsoft is one of the major customers of GB200 chips, and the slowdown in the growth rate of its capital expenditures will have a negative impact on the supply chain.

Secondly, it believes that the cloud computing and AI infrastructure ecosystem is immature, in particular, such as GB200-related network and power infrastructure is still deficient, and the supply chain has not yet kept up with the demand, limiting the deployment of GB200.

At the same time, Morgan Stanley believes that there are also controversies in the market regarding the efficiency aspects of large language models (LLMs), such as the related disagreement between DeepSeek and Microsoft, and that these will continue through 2025, making it difficult for the market to revalue these stocks.

Morgan Stanley also predicts that the cloud computing market growth cycle may continue through the first half of 2025, followed by a fourth-quarter year-over-year growth rate that could fall to single digits.

In March last year, Nvidia unveiled its next-generation Blackwell GPU architecture for AI models, as well as the GB200, a next-generation AI accelerator based on this architecture.