It appears NVIDIA’s Blackwell GPUs have fallen victim to production competitors, as the company now expects smaller units, reaching 500K in 2024.
NVIDIA’s “design flaw” rumors at Blackwell have begun to gain traction as Team Green is expected to delay the schedule by nearly a quarter, with smaller units expected to be produced in 2024
Reports that NVIDIA’s Blackwell products have allegedly hit a snag in mass production are circulating all over the internet, with analysts claiming the issue is a significant one for the company. In our previous coverage on this topicNVIDIA responded to us, claiming that Blackwell sampling has already started, but it seems that the situation is a bit different as new reports claim that NVIDIA has been affected by Blackwell’s design flaws, which is why the company is expected to witness a drop in production volumes this year.
According to JP MorganNVIDIA’s GB200 “Blackwell” AI servers are expected to see annual production of 500,000 units from Team Green, down from the 600,000 units previously planned. The financial firm says design flaws are at play in the shortfall, and that NVIDIA is currently making engineering design changes, a process that will also take more time and money. This could push the launch of Blackwell’s AI servers to the first quarter of 2025, which is about four months earlier than previously planned.
There is uncertainty surrounding Blackwell and its design flaw, but given that Team Green plays a vital role among the major tech giants and that NVIDIA’s dominance is currently unchallenged, the delay won’t have much of an effect, as Bernstein analyst Stacy Rasgon reiterates.
It remains clear that demand levels continue to increase, with all major hyperscalers continuing to increase their capex outlook. Nvidia’s competitive window is so large right now that we don’t think a three-month delay will cause significant share shifts.
– Stacy Rasgon via Reuters
Blackwell’s delay could open a window for competitors to offer their next-gen AI products, but companies like AMD and Intel won’t have enough time given that their respective next-gen AI architectures are scheduled to launch in mid-2025. With Blackwell witnessing a delay, it will create a compounding effect, meaning that the deployment of large AI clusters by companies like Microsoft won’t be possible until mid-2025, creating more problems for those involved in the AI race.