The GPU market is built on a broken foundation

More than a month after their reveal at , I think it’s fair to say NVIDIA’s are a disappointment. Between , the company’s and minimal performance gains over the 40 series, this might be one of NVIDIA’s worst releases in recent memory. But the worst part of it all has been watching history repeat itself.

Looking back at the last few weeks, I’m reminded of an article I . The short of it was I bought an in August of that year, or little more than two years before NVIDIA announced the GeForce 40 series. Plenty of readers rightfully pointed out that I was wrong about the effect the crypto market would have on future pricing, but I think my underlying reasoning was sound. I bought the 3070 when I did because I had a feeling NVIDIA’s new GPUs would cost more than their 30 series counterparts and that finding a new card at release for a reasonable price would be nearly impossible. And I was right: 40 series cards were difficult to find at launch.

Here we are two years later at the start of a new GPU generation, and we’re doing this all over again. If you’re just a regular person looking to upgrade your gaming PC, not only is it impossible to find a 50 series GPU in stock anywhere, but nearly every single model is priced far above NVIDIA’s suggested price. There’s no pandemic to blame this time, so what gives?

The problem is that trends during COVID set a precedent. Since 2020, some people have been willing to pay whatever it costs to buy NVIDIA new cards at launch. The difference now is that rather than give scalpers the chance to profit on that behavior, NVIDIA’s partners have decided to take that money for themselves. The end result is that the company’s GPUs are now marked up as high as the market will tolerate, and they will continue to carry outrageous prices as long as supply is limited.

It’s time we admit the business model at the heart of the GPU industry is broken. For the uninitiated, AIBs (or add-in board partners) like ASUS, Gigabyte, XFX, and Zotac produce the majority of GPUs you can buy from NVIDIA and AMD. In the past, this model led to differentiation between cards in the same tier. For instance, there was a time in the mid-aughts where you could spend extra to buy a GeForce 8800 GT with 1GB of VRAM instead of the 512GB specified by NVIDIA. However, those days are long gone.

Modern AIB GPUs can feature slight overclocks or an extra fan but fundamentally they’re all the same product. Whether you buy a 5070 directly from NVIDIA or one of its partners, it will still be a 5070. At best, you might see a few percentage point difference in performance if you pick a model with an overclock or additional cooling, and yet pricing can vary dramatically between different AIBs and even within one company’s lineup.

Digital Foundry’s said it best in one of the outlet’s recent DF Direct episodes. “The AIB model seems, for a lot of cards, like some weird middle man that purely exists to extract more value from the consumer, and doesn’t offer valid differentiation for the price increase.”

Just look at the . According to , pricing starts at $749. However, retailer websites tell a different story. If you , for instance, there’s not a single 5070 Ti listed for $749. The most affordable one comes in at $799, with every other option priced above $830. There’s even one listed at $920. Needless to say, the 5070 Ti is a very different (and worse) value proposition at $900 than it is at $750.

Newegg briefly the Trump administration’s recent tariffs on Chinese imports for the inflated price of NVIDIA’s new 50 series GPU, and that would have been a plausible explanation if the cost of those cards went up by only 10 percent.

NVIDIA seems uninterested in tackling the problem in a meaningful way. So far, the best solution it has offered is a that gives an unspecified number of US customers the chance to buy a 5090 or 5080 directly from the company. Provided they deliver the performance the company is promising, AMD’s new will put pressure on NVIDIA to price the 5070 and 5070 Ti more aggressively, but it remains to be seen if AMD will manage to keep the cards in stock and at their suggested retail price.

Pricing for this generation’s GPUs will normalize eventually, but unless there’s a fundamental change to how the industry does business, you can bet all of this will happen again. Just because the industry has always done things this way doesn’t mean we should settle for business as usual — especially as the status quo gets worse every launch cycle.

If you buy something through a link in this article, we may earn commission.