Why The Rumored Delay Of NVIDIA's Blackwell GPUs is Unlikely To Be A Deal Killer
The rumors of a delay and design flaw in the upcoming Blackwell GPUs are unlikely to reveal a relevant problem at NVIDIA.
This is the author’s opinion only, not financial advice, and is intended for entertainment purposes only. The author holds a beneficial long position in NVIDIA Corp.
On August 2, online tech publication The Information published a report citing alleged insiders claiming that NVIDIA’s Blackwell GPU shipments will be delayed by at least 3 months due to design flaws. Unfortunately, more information is kept behind a significant paywall. In the following, I would like to discuss why these rumors, even if partially true, may not be relevant to NVIDIA’s future business success.
First of all, there is the question of which Blackwell GPUs we are talking about. The informed reader will of course immediately think of the GB100 AI GPU that NVIDIA plans to release later this year as a single chip or as a dual chip in the form of the GB200 superchip to succeed the incredibly successful Hopper GPUs. Since NVIDIA’s future sales are largely dependent on these GPUs, a delay could actually become a problem for NVIDIA.
However, other next-generation NVIDIA GPUs such as the RTX50 series are sometimes also referred to as “Blackwell” generation. There have already been rumors of a delay in the delivery of the RTX50 “Blackwell” GPUs, with production of these chips being postponed in order to create more capacity for the GB100 and GB200. The Information's report that the delay also affects Microsoft or Amazon would not contradict that. Microsoft Azure and Amazon Web Services also offer NVIDIA RTX virtual workstations. Therefore, NVIDIA would also have to inform these customers of a delay in the delivery of these chips. In terms of sales, however, this delay should not play a major role, as the GB100 and GB200 are NVIDIA's highly expected future cash cows, not the RTX50 series.
Despite this, NVIDIA CEO Jensen Huang said at Siggraph recently, “This week we are sampling we are sending out engineering samples of Blackwell all over the world they are under people’s chairs right now”. I very much doubt that Jensen Huang would make such statements if there were still serious problems with the GB100 and GB200. Furthermore, an NVIDIA spokesperson has already reassured on August 3 that production of Blackwell GPUs will ramp up in the second half of 2024.
But even if there is a design flaw, it is most likely the interconnect between two Blackwell GB100s to a GB200. This would mean that the individual Blackwell GPUs are not affected. It has already been reported here that problems with this interconnect could lead to production delays of two weeks. However, such problems are not uncommon when starting production of a new, highly complex product like the GB200, and are usually resolved in a short period of time, resulting in an improved product.
For the above reasons, I think it’s unlikely that The Information’s report uncovered a relevant problem at NVIDIA. No complex product has ever been rolled out at scale without at least some minor problems during development or production. Main Street and Wall Street are looking closely at the upcoming Blackwell GPUs as the "engine of the new industrial revolution" - any rumor here guarantees attention and clicks. However, in today's nervous market, such rumors could put short-term pressure on the stock.
Follow me on X for frequent updates (@chaotropy).
