Why is it important: Not surprisingly, despite the very high price of the card, the RTX 4090 has been depleted of stock at most retailers, indicating that we are in a different situation where demand far exceeds supply. But if the new report is accurate, there could be even fewer AD102 GPUs available as Nvidia spreads production to more profitable Hopper-based H100 enterprise GPUs.
A quick look at Newegg for the RTX 4090 reveals nothing but “out of stock” notifications, which were a depressingly familiar sight in the midst of the graphics card crisis. The MSRP of $1,499 is high, but the sheer performance of the current Lovelace flagship has made it appealing to those who want the best of the best.
According to My Drivers, the RTX 4090 shortage may be exacerbated by Nvidia commissioning Taiwan Semiconductor Manufacturing Co (TSMC) to move some production from the AD102 GPU that powers the RTX 4090 to the H100 compute processors. They both use the same 4N node, so switching production shouldn’t be a big deal for TSMC.
Why would Nvidia take such a step? Like so many other things in life, a lot of it probably comes down to money. The RTX 4090 may have high margins, but it can’t compete with the H100 GPU; the $30,000+ SXM variant includes 16,896 FP32 CUDA cores, 528 Tensor cores, and 80GB of HBM3 memory connected using a 5120-bit bus. H100-based products sell for tens of thousands of dollars, and companies often buy them in bulk for server farms and exascale computers, which means healthier profit margins for Nvidia.
Nvidia probably needs more money. It had a terrible second quarter due to falling gaming revenue and declining consumer demand due to skyrocketing inflation. It’s also facing a huge bill to “launch” the RTX 4080 12GB, especially since the green team is reportedly paying most of the bills from AIB partners who have to repack and rename the card.
But the biggest financial blow Nvidia could take could come from the US government. A few weeks ago, US officials ordered Nvidia and AMD to stop selling their high-end AI-focused GPUs to China, which is expected to cost the former company $400 million.
As with all unverified reports, take this with a healthy dose of skepticism. But Nvidia may see moving AD102 production as the best course of action, especially with AD103-based RTX 4080 16GB cards and AD104-based RTX 4080 12GB cards, which should help meet some of the demand.
Thanks, Tom’s equipment