Friday, February 23rd 2024
NVIDIA Expects Upcoming Blackwell GPU Generation to be Capacity-Constrained
NVIDIA is anticipating supply issues for its upcoming Blackwell GPUs, which are expected to significantly improve artificial intelligence compute performance. "We expect our next-generation products to be supply constrained as demand far exceeds supply," said Colette Kress, NVIDIA's chief financial officer, during a recent earnings call. This prediction of scarcity comes just days after an analyst noted much shorter lead times for NVIDIA's current flagship Hopper-based H100 GPUs tailored to AI and high-performance computing. The eagerly anticipated Blackwell architecture and B100 GPUs built on it promise major leaps in capability—likely spurring NVIDIA's existing customers to place pre-orders already. With skyrocketing demand in the red-hot AI compute market, NVIDIA appears poised to capitalize on the insatiable appetite for ever-greater processing power.
However, the scarcity of NVIDIA's products may present an excellent opportunity for significant rivals like AMD and Intel. If both companies can offer a product that could beat NVIDIA's current H100 and provide a suitable software stack, customers would be willing to jump to their offerings and not wait many months for the anticipated high lead times. Intel is preparing the next-generation Gaudi 3 and working on the Falcon Shores accelerator for AI and HPC. AMD is shipping its Instinct MI300 accelerator, a highly competitive product, while already working on the MI400 generation. It remains to be seen if AI companies will begin the adoption of non-NVIDIA hardware or if they will remain a loyal customer and agree to the higher lead times of the new Blackwell generation. However, capacity constrain should only be a problem at launch, where the availability should improve from quarter to quarter. As TSMC improves CoWoS packaging capacity and 3 nm production, NVIDIA's allocation of the 3 nm wafers will likely improve over time as the company moves its priority from H100 to B100.
Sources:
Q4 Earning Call Transcript, via Tom's Hardware
However, the scarcity of NVIDIA's products may present an excellent opportunity for significant rivals like AMD and Intel. If both companies can offer a product that could beat NVIDIA's current H100 and provide a suitable software stack, customers would be willing to jump to their offerings and not wait many months for the anticipated high lead times. Intel is preparing the next-generation Gaudi 3 and working on the Falcon Shores accelerator for AI and HPC. AMD is shipping its Instinct MI300 accelerator, a highly competitive product, while already working on the MI400 generation. It remains to be seen if AI companies will begin the adoption of non-NVIDIA hardware or if they will remain a loyal customer and agree to the higher lead times of the new Blackwell generation. However, capacity constrain should only be a problem at launch, where the availability should improve from quarter to quarter. As TSMC improves CoWoS packaging capacity and 3 nm production, NVIDIA's allocation of the 3 nm wafers will likely improve over time as the company moves its priority from H100 to B100.
64 Comments on NVIDIA Expects Upcoming Blackwell GPU Generation to be Capacity-Constrained
"we won't be able to make enough of these, they'll fly off shelves", because of course they've said some version of that, and it's probably true.
RDNA3 was a learning process for them with regard to MCM packaging and the wins/losses that come from that approach and RDNA4 is supposed to essentially be a bug-fixed and much more optimized version of RDNA3.
RDNA5 is supposed to be six to nine months behind RDNA4 which seems to indicate that RDNA4 is more of a half-generation GPU series, more like a "super" release but targeted toward the midrange or lower cards.It isn't wishful thinking as much as it's a Twitter/Discord circle-jerk by people that went to the YouTube school of Engineering.DLSS, FSR, and XeSS's days are numbered.
Microsoft and the Khronos Group are sick of vendors doing proprietary shit and have some smart people working on vendor-independent upscaling.
Edit: before any of you start pounding out an empassioned response to tell me that DLSS is hardware accellerated, whatever DirectX and Vulkan upscaling standards come out of this will allow for optional hardware accelleration either on the GPU or CPU. Nvidia/AMD/Intel can write their drivers so the upscaling tech executes using the same hardware features that they're currently using. That's how it went down with multitexturing, shaders, deferred rendering, and so on.
I also wouldn’t rule out shrinks just yet. Sure, we may be approaching limits of silicon as a material, but other options exist. Graphene and synthetic diamonds have shown some potential here.
Telling NVIDIA to care about gamers, my friends they are not a social services company.
It hurts yes, gamers and crypto have been feeding the monster.... now we have this.
We just need to have some solid alternatives from AMD and Intel.
The obvious fix for this is to get off PC gaming. But as that won't happen, it's cloud or 30k for a GPU. So pick one of the three.
What's sad is, suckers will still buy no matter the price, therefore keeping nGreedia happy and the stocks up.
Let's hope Intel and AMD will bring some competition, but it's hardly believable...Fixed.
Remember the time when video cards had 2 two chips? One for 2D graphics and one for 3D one?
But of course, you just "make" them for like, "really cheap" and then "charge thousands" because it's greed and not because they spent multiple billions on R&D and have actual constraints involving third parties, technology and at this scale, even the concept of physics itself. Money just solves (absolves) everything!
How long the supply will be an issue after launch remains to be seen though.
making statements like this is stock manipulation 101 and super illegal
The TSMC N3 node is in extreme demand, and the processors Nvidia has are easily amongst the most advanced using this node, which means that yields are not exactly perfect. Everything they said is true.
RDNA4 may only be mid-range but the 8700XT class gpu is said to be faster than 7900XT yet under $500.