Recent statements from AI companies indicate that the AI boom has increased demand for GPUs to the point where some are reporting shortages. Nvidia claims that its supply chains have not completely failed to meet demand, but have encountered minor hurdles.
Artificial intelligence can create a video card crisis
Microsoft, OpenAI and other companies talk about the steps they have taken to reduce the shortage of Nvidia H100 and A100 enterprise GPUs used for AI workloads. Microsoft imposes wage caps on its employees, and the Quora CEO says hardware shortages hide the true potential of AI applications. Elon Musk even joked that enterprise GPUs are now harder to get than drugs. This “joke” is important as the Dojo supercomputer was built for Tesla.
On the other hand, Nvidia explained that production problems with GPUs are mostly related to another component, not the main processing unit. Currently, the problem stems from TSMC’s CoWoS packaging technology. Nvidia promises that supply will be much stronger in the second half of 2023.
It’s unclear how far this will affect the consumer GPU market, but the artificial intelligence boom has already caused Nvidia to divert some production resources from GeForce RTX 4090 graphics cards to H100s. While comments from AI companies suggest that it will take time for attention to shift to AMD hardware, a post by hacker George Hotz indicates that his company, Comma, has purchased Radeon 7900 XTX cards for this purpose.
Earlier this year, groups that previously used GPUs for mining began attempts to shift their infrastructure towards AI workloads to recoup their investment after the crypto winter. While this shift may prove fruitful for some, the competition will be tougher as big players like Microsoft, Amazon Web Services and OpenAI are involved.