Nvidia is improving it’s AI GPU delivery times

Consumer cards might not be selling like hto cakes but AI GPUs certainly are

Sebastian Kozlowski

Share this article...

• Follow us

There have been reports that Nvidia’s AI GPUs are taking a lot less time to arrive at its customers than before. As we’ve seen the lack of RTX 4090 stock, the AI boom is not slowing as we’ve seen the need for these cards growing and trying to gain every processor available for those large language models.

Now according to MyDrivers (translated by Google), UBS analysts’ memo provides an update to investors saying that the delivery speed of AI Nvidia GPUs has accelerated at the end of the last year. Shortening the wait time from 8-11 months to only 3-4 months. This shows a great improvement to the supply chain for the high volume orders, as unlike consumers, they’re not just buying one GPU for a system so it is a different challenge entirely.

Best GPU deals

Nvidia’s limits and production

One of the main reasons supposedly is that TSMC has managed to increase the production capacity of its CoWoS packaging, in fact, it is aiming to double what it has to offer, massively boosting interest in it. It also allows the likes of Nvidia to produce more GPUs as well as Samsung Foundry to manage the HGM department and resolve some of the memory limits.

But that is not the only reason for the freeing of its processors. Toward the end of 2023, the US banned AI chip exports to China, cutting down a great deal of sales Nvidia might have had into the country. Pivoting those away allowed them to go towards a smaller number of customers. Even with the specific 4090D created for the market, it may not reach the same level of sales.

Which leads to the next problem of limited demand. With a slow expansion, and fewer markets to access, then fewer GPUs will be sold in general. Especially as competitors, even domestic ones like Huawei will be looking to fill the gap, Nvidia may not have such great growth moving forward.