Shortage of AI GPUs could affect graphics card market
In recent years, the cryptocurrency boom has caused a significant increase in the prices of PC graphics cards throughout 2020 and 2021, and even today prices are not where many users would like them to be.
Photo: Allbreaknews.com
Now, with the current frenzy around artificial intelligence (AI), the industry is worried that it could face a similar effect. Companies report a shortage of GPUs, but Nvidia says the situation is more complex than many realize.
Recent statements from AI companies indicate that the boom in this technology is increasing demand for GPUs to the point where some companies are reporting shortages. Nvidia claims that its supply chains are not totally unable to meet demand, but are facing minor obstacles, allaying fears to some extent.
Companies such as Microsoft and OpenAI mention measures they have had to take to mitigate shortages of GPUs for AI, such as Nvidia's H100 and A100 models, used for AI workloads. Microsoft is limiting the rate at which employees can use them, while Quora's CEO claims that the shortage of hardware is masking the real potential of AI applications. Elon Musk himself joked that enterprise GPUs are currently harder to acquire than drugs.
According to Nvidia's explanation, the bottleneck in GPU production often doesn't involve the main processing unit, but some other component on the circuit board. In the current case, the problem is related to packaging, which the company is currently addressing with TSMC's 3D chip stacking technology (CoWoS – chip-on-wafer-on-substrate). Nvidia promises that supply will be much more robust in the second half of 2023.
It's not yet clear how much this situation might affect the consumer GPU market, but the AI boom has already caused Nvidia to redirect some production resources from GeForce RTX 4090 graphics cards to H100s GPUs. Although comments from AI companies suggest that shifting attention to AMD hardware would take time, a post by prolific hacker George Hotz indicates that his company, “comma”, is acquiring Radeon 7900 XTX cards for this purpose.
Although Nvidia currently dominates the AI market, analysis by TechSpot suggests that AMD could certainly make significant moves in the sector. AMD's main advantage in AI is its leadership in CPUs for Windows PCs, which incorporate AI blocks.
Earlier this year, groups that previously used GPUs for mining began trying to switch their infrastructure to AI workloads in order to recoup investments after the cryptocurrency winter. While this shift may be fruitful for some, it will probably be more challenging than mining, due to competition from big players such as Microsoft, Amazon Web Services and OpenAI.
