NVIDIA GPUs are apparently having their “fine wine” era — but not in the way PC gamers usually mean it.
For years, GPU fans used that phrase when an older graphics card suddenly performed better after driver updates, game optimisations, or improved software support. This time, NVIDIA CEO Jensen Huang is using the idea for something very different: price.
According to comments highlighted by Wccftech, Huang said GPU consumption is “going through the roof”, and even GPUs sold four or five years ago are now climbing in price faster than “good wine”. He even compared buying NVIDIA GPUs to investing in art.
That is a wild statement, but honestly, it tracks with where the tech industry is right now. AI companies, cloud providers, and data centres are still hungry for compute, and GPUs remain the main engine behind a lot of modern AI workloads. CPUs are also seeing stronger demand, but for AI training and inference, NVIDIA accelerators are still the hot item.
Why older GPUs are still valuable
Normally, old hardware drops in value as new generations arrive. Gamers know the pain: you buy a GPU, a newer model launches, and suddenly your card feels less exciting.
But AI hardware does not age the same way as consumer gaming GPUs. Older accelerator-class NVIDIA chips can still be useful if the software stack keeps improving and if companies can actually get access to them. The source specifically mentions cards such as Hopper H100s, H200s, L40S, and even the previous-generation A100s as examples of older GPUs seeing stronger demand.
CoreWeave’s CEO also reportedly said demand for these older GPUs is accelerating, with prices for those parts becoming more expensive compared to the previous quarter. The company says it remains largely sold out across its fleet capacity.
In simple terms: if a data centre cannot get the newest GPU, an older NVIDIA accelerator is still better than having no GPU at all.
This matters for Malaysia and SEA too
For Malaysian gamers, this does not mean your old RTX card suddenly becomes a retirement fund overnight. The article is mainly talking about data centre and AI-focused GPUs, not your average gaming rig sitting under the desk.
But the bigger trend still affects us. When AI companies absorb wafer supply, GPU allocation, DRAM, and cloud compute capacity, the entire hardware chain gets tighter. That can influence everything from workstation pricing to server rental costs, and eventually even the price of gaming laptops, creator PCs, and local AI infrastructure.
For SEA startups, universities, game studios, and content teams experimenting with AI tools, this also means compute may stay expensive. Renting AI cloud capacity could remain painful if providers are sold out or forced to pay more for older-but-still-useful GPUs.
For PC builders in Malaysia, the practical takeaway is simple: do not assume the GPU market will behave like the old days, where last-gen hardware automatically becomes cheap after a new launch. AI demand has changed the equation.
The GPU market is not normal anymore
The source also points to wider pressure across the semiconductor supply chain, including wafers, GPUs, CPUs, DRAM, and cloud compute capacity. Basically, everyone wants more hardware than suppliers can comfortably provide.
That is why this “fine wine” comment hits differently. It is not about gamers getting better frame rates through driver magic. It is about old NVIDIA GPUs staying valuable because AI demand is still outpacing supply.
Good for NVIDIA, no doubt. For the rest of us? Better start watching GPU prices like we watch Steam sales.
Source: Wccftech Gaming