NVIDIA RTX 5090 might end up a tamer GPU if AI boom continues - but that could work out well

NVIDIA's RTX 5090, or whatever the next-gen flagship GPU ends up being called, might not be as much of a leap forward as previously thought.

This is one of the latest bits of chatter to come via the YouTube rumor mill, specifically leaker Moore's Law is Dead who is forthcoming with a lot of material on future GPUs and CPUs.

In this case, Moore's Law is Dead was talking about the subject of whether NVIDIA will cancel its high-end GPUs (following suit with AMD, which at this point is heavily rumored to be pushing aside top-end RDNA 4 offerings, seemingly to focus on RDNA 5 and making that better).

The YouTube leaker believes that NVIDIA will produce an RTX 5090 for next-gen Blackwell, but that this might be the last such offering - so the RTX 6000 series might see this top dog GPU abandoned.

Moore's Law is Dead argues that if the AI bubble continues at the pace it has been inflating, and even accelerates over the next six months - which given the fervor around AI right now is hardly inconceivable - then NVIDIA might cancel the very top Blackwell configuration.

That doesn't mean there won't be an RTX 5090 - just that NVIDIA will switch away from a more powerful planned incarnation of the Blackwell flagship that's in the cards right now, to a lesser spec.

Moore's Law is Dead describes the more powerful take on the RTX 5090 as "pretty damn huge and expensive" compared to the second option which is "quite smaller" the leaker reckons. So, it does sound like a fair old difference.

And we certainly wouldn't bet against the AI market picking up even more momentum than it's currently experiencing.

Show me the money

This kind of switch from NVIDIA, to favor AI graphics cards over GeForce gaming models, makes sense because of the profit to be realized from the former.

As we've just seen in a tweet highlighted by Tom's Hardware, NVIDIA is apparently selling H100 GPUs for approaching $30,000 when it costs only a tenth of that (or just over) to make them (that's a rough calculation, but still, you get the idea).

While NVIDIA also takes a good chunk of a premium on its high-end gaming cards, the profit obviously can't hold a candle to the kind of level that can be achieved selling to the AI sector.

So, the future of GeForce gaming GPUs appears to be in the hands of the AI industry, and is dependent on the exact demand fired up over the rest of this year (and into next).

Tom's tells us that NVIDIA's AI acceleration boards are already sold through entirely for the remainder of 2023, and the market is forecast to be worth $150 billion by 2027. The gravity of the potential profits to be mined (no pun intended) here can't be underestimated.

Another reason Team Green could take its foot off the GeForce accelerator is the aforementioned rumor that AMD isn't even going to try and compete in this space, anyway (and Intel certainly won't).

The only good thing about NVIDIA taking the above-mentioned lesser spec route for the RTX 5090 is that it'd make the flagship GPU a fair bit less expensive. Could we even be looking at a $1,000 flagship graphics card for Blackwell, perhaps?

Well, that's an optimistic best-case scenario, but you get the point - a more reined-in RTX 5090 may not be such a bad thing after all. At least in theory, though with NVIDIA's pricing strategy, we can never be sure where price tags might end up...