As I have opined in other threads, I see nothing but down side for nVidia's GPU business. The best they can do is to not slip too far. In the AI business, nVidia has no equal. I see good things out of AMD and a couple of others but nVidia gets 100% of data centre wins. What's more, the cost of AI is energy, not initial investment. It appears nVidia will be able to charge a premium for some time to come. If we were talking about GPUs, I would bet on AMD. Not to take the lead but to substantially shrink the gap. With AI, the only sensible bet seems to be nVidia. Meanwhile, in China, they have some reasonably powerful chips. They are off the performance mark but that might still be OK. The problem is they require way more power to do the same thing. They cannot win AI until they have a competitive semiconductor process. The Chinese gap will continue to widen.
Here's my short take: Ever since HD 5870, AMD has not been a leader in GPU anything for those who can pay. If you wanted the best, Nvidia it is. If you have a productivity load, Nvidia and CUDA it is. If you want the best driver support, Nvidia it is. Best software stack? Nvidia by a mile. Best game features? Nvidia. Every gamer forum I go through, AMD is looked at as either a value for money for certain SKUs or a perpetual second best when an Nvidia SKU is out of stock. Nvidia spends serious money on their software ecosystem. AMD? Not so much. They go the open source route, and that has not been a winning formula for success. Looking forward, AMD is going to have to come out as a value leader by a large margin if they want to peel away more from Nvidia. They are going to have to sacrifice margin for market share. And they are going to have to sink a lot of money on their software stack. That's an absolute necessity. CUDA didn't come cheap or easily. With that in mind, it's not going to help them financially, but mindshare costs a lot to gain. For gamers, it's a drivers and cost per frame thing. For creators, AMD has a LOT of work to do. Ever since AMD acquired ATI they have struggled to make a dent in Nvidia's dominance. If it was just about any other company besides Nvidia, AMD would be on top because of Lisa Su, but Jensen Huang is a damn brilliant CEO. It's hard to outmaneuver a market leader with someone like him at the helm. The fact that both of them are related? Chef's kiss. Now with all that out of the way, let's get to "I see nothing but down side for nVidia's GPU business". AMD already dominates the iGPU side. Game consoles? Check. You think that continued dominance will yield more than it already is? What are you seeing that I am not?
Reading more about DeepSeek, this is quite the wrench in many peoples' plans. Of course NVDA and TSMC are affected due to a potential sea change of AI GPU orders from the big players... BUT... this might also have the effect of smaller players snapping up cancelled orders. I am not sure I see this as a crap your pants NVDA sell off moment. This is perhaps a watershed moment for a sort of democratization of AI training. The world still needs computational power. NVIDA is the best, at least for the near future. To me, the fact that this is Chinese software is very interesting. It might play to our current administration's AI aspirations. We might have greatly accelerated this AI arms race with this one piece of code. It's going to be extremely interesting to see how this all plays out. I would LOVE to be a fly on the wall near the Sam Altman's of the world right now!
The new Trump/Altman AI project with a half trillion dollars of funding should be a massive boost to nVidia. It will involve nearly no R&D. It's just build and sell. Epic.
As an investor, I am literally banging my head against the wall with how badly NVDA is botching the RTX 5 series roll out. A new chapter in the saga: The door is wide open for you, AMD. Don't screw it all up like you have a tendency to with your GPUs!
It looks like 5000 series GPUs won't be available *in quantity* until Q3. It appears there are some issues nVidia need to sort out. So far, I believe AMD is doing everything right. The GPU race won't be like Intel/AMD where AMD takes over the majority of the volume. This will be a step from 20% market share to 40%, best case. Zero chance AMD can take the GPU lead, even if nVidia stumbles for a while. What's more, even if AMD eats a bit of nVidia market share, that will matter to AMD but isn't that big of a problem for nVidia. nVidia is quickly becoming an AI company where they own the market. After using DeepSeek for a couple of months, I'm not sure what to think of where it fits. I consider DeepSeek second best but not by much and their claim of using an order of magnitude less compute is haunting.
Here is Tom's take on the nVidia 5000 series. The 5090 has 33% more resource (various cores, memory bandwidth, etc.) than the 4090 while bringing about a 30% performance boost in raw fps. It looks to me like there is little to no generational uplift going on here. I would have expected an efficiency boost. The parts use the same TSMC 4N process. This is a brute force hardware step forward. The biggest improvement is on the software side. The party trick of the 5090 is that it will legitimately game at 4K. It doesn't matter how it got there. LTT did a 4090/5090 comparison that showed they could game comfortably at 4K with high frame rates. I don't see the $2000 price as a problem for the 5090. I think the 600 Watt TPU is more of a problem. Even the wild power consumption won't be a game stopper. I suspect the 5090 will sell into a similar size market as the 4090 did.
There are some pretty strong rumours that nVidia is under a product recall for 5000 series parts. If you hold nVidia, you may not like the next few days.
nVidia is currently in spin control. Link: https://www.theverge.com/news/618748/nvidia-admits-the-rtx-5080-is-affecte NVIDIA's Response Below: “Upon further investigation, we’ve identified that an early production build of GeForce RTX 5080 GPUs were also affected by the same issue*.* Affected consumers can contact the board manufacturer for a replacement*,” Nvidia GeForce global PR director Ben Berraondo tells The Verge.* In response to The Verge’s questions, Berraondo adds that “no other Nvidia GPUs have been affected” — we specifically asked about the upcoming RTX 5070, and he says it’s not affected either. Nor should any cards be affected that were produced more recently: “The production anomaly has been corrected,” he says. In case you’re wondering, he also told us that Nvidia was not aware of these issues before it launched these GPUs. Here's NVIDIA's Full Amended Statement: We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D, RTX 5080, and 5070 Ti GPUs which have one fewer ROP than specified. The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected. [Channeling WXYZ] Tom's take on this. That's the last official statement from nVidia but there are field reports that some cards are missing 8 ROPS, not 1, and we also have field reports that retail are being asked to return entire runs of RTX 50 laptops. If there were one or two reports of laptop returns, it would be a lot easier to dismiss this rumor. GPUs are a religion. It's been clear since late December that RDNA 4 is going to significantly close the gap between AMD and nVidia. Saying so online makes an enemy out of every nVidia GPU owner. They declared the 5000 series superior to RDNA 4 before RDNA 4 was ever tested. So... somebody is lying. In this case, I think nearly everyone is lying. nVidia has a problem. That's a fact. nVidia has a recall and they have also said they are going to sue the people who have said there is a recall because there is no recall. I suspect this is going to be one of those events that hindsight will show could have been easily handled with a few thousand replacement parts. From there, they could carry on and it would be quickly forgotten. It would be something quickly forgotten but I believe their entrenchment in a lie will make things much worse. Lastly, the 5000 series has some problems but it is still an extremely impressive product. Most of the problems will be fixed over time, like every other GPU release. Some of the problems will be fixed when the 6000 series and that is said to be sooner than later (maybe late 2025 with another paper launch). nVidia has a shot at spinning this because they have delivered so few parts to consumers. The 5000 series was the least substantive paper launch we've seen in 25 years so we are going to see if they can put this genie back in the bottle. If they could spin up production and flood the market with correctly configured parts, that would help but that's going to cut into their AI production which is their bread and butter. We are about to see what nVidia is made of: take care of customers to preserve corporate image or go for maximum profit and keep all wafer production for AI parts.
