NVIDIA and the drop in shipments: new scenarios for the semiconductor sector | Cpu-z | What is hardware | Gpu vs cpu | Turtles AI

NVIDIA and the drop in shipments: new scenarios for the semiconductor sector
Morgan Stanley Forecasts Downgrade GB200 Volume as DeepSeek, US Tariffs Reshuffle Market Dynamics
Editorial Team5 February 2025

 

Morgan Stanley has revised down NVIDIA’s GB200 shipment estimates for 2025, with significant supply chain impacts. The slowdown in the cloud computing market and Trump’s tariff tensions on TSMC exacerbate uncertainty in the semiconductor industry, while DeepSeek threatens NVIDIA’s dominance with more efficient AI solutions.

Key points:

  • NVIDIA forecast reduction: Morgan Stanley lowers estimates on GB200 NVL 72 shipments from 30,000-35,000 to 20,000-25,000 units.
  • DeepSeek factor: The cheaper and more efficient R1 model imposes new competitive dynamics in the AI industry.
  • Slowdown in cloud computing: The industry’s CapEx could decelerate with single-digit growth by late 2025.
  • U.S. duties on TSMC: Trump-imposed tariffs on Taiwanese chips could push TSMC to raise prices more than 15 percent.


Morgan Stanley recently updated its forecast for NVIDIA’s GB200 NVL 72 shipment volume, drastically reducing it from the previous 30,000-35,000 units to a range of 20,000 to 25,000 units, with the risk of a further drop below this threshold in the worst-case scenario. The financial institution’s new report highlights how the entire GB200 supply chain could suffer an economic backlash of between $30 billion and $35 billion, reflecting the slowdown in the cloud computing market and a changed competitive environment. According to analysts, the cloud industry’s expansion cycle is expected to continue through the first half of 2025, but the annual growth rate could drop to single-digit values in the fourth quarter of that year. On the competitive front, Chinese giant DeepSeek has destabilized the AI landscape with its R1 model, which is capable of achieving performance comparable to OpenAI’s O1 model with significantly reduced costs, estimated at about 1/50th of the typical investment required for training advanced LLMs. Although the financial analysis suggests that DeepSeek has incurred a CapEx of $1.6 billion and operating costs close to $944 million, the impact of its efficient approach has already been felt by NVIDIA, contributing to the downsizing of the GB200 NVL 72 shipment forecast. Factors fueling uncertainty in the semiconductor industry include U.S. trade policies, particularly President Trump’s desired tariffs on Taiwanese chips, which could go into effect on Feb. 18. These measures are intended to incentivize TSMC to shift more of its production to the United States, but they risk having the opposite effect, with increased production costs that would result in a list price increase for advanced processes, initially expected to be between 5 percent and 10 percent, but now could exceed 15 percent.

The combination of these elements is redefining the dynamics of the global semiconductor market, amid macroeconomic uncertainties, technological competition, and changes in the strategy of key players in the industry.