Artificial intelligence GPU shortage has begun
TSMC acknowledged that heavy demand from the AI industry has severely hampered supply chains, saying the shortage of AI GPUs from Nvidia and other companies could last up to 1.5 years.
TSMC Chairman Mark Liu admitted that there has been a huge increase in orders from Nvidia and its partners, mainly due to the rise of GenAI (generative artificial intelligence) development, which requires massive computing capabilities. TSMC stated that the bottleneck in CoWoS packaging continues and that the company was caught unprepared.
TSMC is witnessing three times the normal orders, especially for CoWoS packaging services, as this process accounts for a large share in the production of AI GPUs. However, TSMC stated that they are working to expand the facilities to meet the demand from the artificial intelligence industry and that they hope to double production by 2024. As announced by the Taiwanese giant, this does not seem to be enough for now as the shortage could continue for a long time. It is said that the famine may continue until at least 2025.
Recently, both AMD and Nvidia have stated that their AI-based revenues could reach hundreds of billions of dollars. However, this prediction is overly optimistic. Because such targets usually do not take into account the production shortage. However, eventually the production shortage will end and demands will be met. On Nvidia’s side, things are better than AMD. The company has a very right product at a very right time: H100. However, Nvidia plans to sell 1.5 to 2 million H100s by 2024.