SemiAnalysis sharply cuts its AI chip demand forecast: Nvidia may face the greatest pressure

On November 16, AI research firm SemiAnalysis sharply lowered its forecasts for AI chips, server shipments, and related capital expenditures from 2028 onward in its latest report. It said that slowing growth in new ARR at frontier AI labs, tighter financing conditions, and delays in U.S. data center construction would collectively constrain computing-capacity expansion in the coming years.

Under the firm’s revised model, global capital expenditures on AI accelerators are projected to reach $758 billion in 2028, down 28% from its previous forecast. They are then projected to fall to $796 billion in 2029 and $923 billion in 2030, down 41% and 40%, respectively, from previous forecasts. The report says the cuts are mainly due to falling chip prices, lower shipment volumes, and reduced assumptions about power consumption per chip in future products.

Among major suppliers, Nvidia is expected to bear the brunt. SemiAnalysis forecasts that Nvidia’s shipments from 2027 to 2030, measured by computing revenue, will be revised down by 22%, 39%, 56%, and 66%, respectively. The firm also lowered its expected average selling prices for Nvidia’s upcoming Feynman series products: Feynman from about $110,000 to about $78,000, and Feynman Ultra from about $175,000 to about $98,000.

The report says Nvidia’s challenges stem not only from a slowdown in overall demand, but also from customers reassessing their priorities around cost, energy efficiency, and financing capacity. Anthropic prefers options such as TPUs, Trainium, and AMD GPUs, while OpenAI is expected to prioritize deploying its in-house ASIC projects. Meanwhile, Nvidia has significant exposure to neocloud customers, which are facing greater financing and execution risks.

View the original: https://www.theblockbeats.info/flash/377610