If you're a PC gamer, upgrading your graphics card is something you're always thinking of. You're either on the verge of buying a new graphics card or planning for an upgrade in the near future, because, let's face it — GPUs don't last as long as they used to. You're one of the lucky few if you're able to get your hands on a current-gen card at reasonable prices. On the other hand, if you're thinking of upgrading when the next-generation cards drop, you might be in for a rude awakening.
The great GPU shortage caused by the crypto boom of 2021-22 still haunts many gamers to this day. But, there might be another GPU shortage waiting for us if the AI wave is anything to go by. Generative AI, large language models (LLM), and neural networks are terms that have seeped into common conversation, and companies have already taken strides to capitalize on the AI boom. Your dream of upgrading your graphics card might just end up being just that.
Enterprises big and small are stockpiling graphics cards
With any revolutionary technology comes droves of players competing to get a larger slice of the proverbial pie. This has already started happening in the world of AI where both big players and emerging startups are doing everything they can to get their hands on GPUs. To power their AI workloads, companies need data center graphics cards like the Nvidia HGX H100. But, due to lingering supply issues and unprecedented demand for AI-optimized GPUs, some companies have been reported to be scooping up gaming graphics cards too.
A report by Gartner estimates that the industry will spend more than $50 billion on AI chips this year, and double that in the next four years. This stupendous growth might be a positive sign for the AI industry, but it rings alarm bells for gamers looking to grab just one decent, affordable graphics card for their personal machine. If manufacturing issues aren't resolved soon, the availability of gaming graphics cards could go from bad to worse in the coming year, as enterprise demand ramps up and server GPUs are hard to come by.
Nvidia isn't concerned about gamers, only profits
Nvidia has a stranglehold on around 60-70% of the skyrocketing demand for AI-focused silicon. This is due to its first-mover advantage, superior GPU performance, and CUDA support. With the company making around $10-$11 billion in a single quarter selling data center GPUs, CEO Jensen Huang's focus has understandably shifted away from the gaming side of the business. The poor value of the RTX 4000 graphics cards might just be the beginning of the end of gamer-friendly practices from Team Green.
After the poor reception of the RTX 2000 series, the subsequent RTX 3000 cards brought prices back down to sane levels, at least on paper. But the introduction of a "90" class card with the RTX 3090 and then the sky-high prices of the RTX 4000 series confirmed Nvidia's new approach of targeting only the whales instead of the average consumer. All marketing regarding affordability and value has vanished and been replaced with hyperbolic statements about luxury cards like the RTX 4090 and even the RTX 4080.
People aren't upgrading their graphics cards as often as they did, because it simply isn't sustainable anymore. PC gaming on the whole is getting way too expensive, pushing many people to buy consoles over gaming PCs. And the flag bearer of this dark era is the graphics card industry, producing products that aren't doing enough to justify the premiums they're charging. Gaming GPU prices will never return to normal again, and Nvidia isn't even concerned.
AMD and others aren't far behind in the AI arms race
If you were hoping that Nvidia was the only one making hay while the sun shines, you've got another thing coming. AMD, the primary GPU competitor to Nvidia, is also ramping up efforts to bolster its AI offerings, specifically the MI300X server GPU. Set to compete with Nvidia's H100 and Intel's Falcon Shores 2 chip, the MI300X will be exceptional for inference workloads, meaning it'll excel at deriving insights from data instead of simply training LLM models.
Microsoft is building custom AI processors, and Google, Amazon, and Meta are using a mix of custom and Nvidia hardware for their AI servers. Historically, AMD's RX series graphics cards haven't been able to put a dent in Nvidia's dominance in the gaming GPU market, owing to poorer ray tracing and upscaling tech. But, they're still the better cards in terms of raw performance, value, and VRAM. If AMD goes the Nvidia way, these advantages might start to disappear starting with the upcoming generation.
Gaming studios have dropped the ball
You'll be all too familiar with the hilariously poor performance of PC ports in recent years. Big-budget titles like Starfield, Hogwarts Legacy, The Last of Us Part 1, and many more landed in a broken and unplayable state on launch. It seems that of late, game studios have stopped bothering to optimize games for multiple platforms and simply put their faith in upscaling to make up for the shortfall. But, this might not be entirely true.
Some reports citing anonymous developers suggest that it's simply becoming impossible to ensure stable performance on PC owing to the infinite hardware combinations at play. This is in stark contrast to optimizing for, say, the PS5 or Xbox Series X where there's a great degree of certainty in terms of the GPU the game communicates with. Other reports suggest a lack of in-depth game engine knowledge on the part of developers due to the ubiquity of Unreal Engine.
Even this isn't a slam dunk of a theory, considering many 2023 games were developed on Frostbite, RE Engine, and proprietary engines from Naughty Dog and Team Ninja. So, it isn't terribly clear what's at play here. Whatever it may be, the fact is that buying a top-tier graphics card will do little to ensure top-tier gameplay if the industry is busy grappling with optimization challenges, unrealistic deadlines, and other unnamed issues.
"Mid-range" cards won't cut it anymore
If you're thinking, "I'll simply buy a cheaper graphics card," you'd better rethink your strategy. Both Nvidia and AMD have slowly but surely redefined what mid-range graphics cards mean. What used to be a $250-$300 price range has now moved significantly up the price ladder to a point where we have to call the $600 RTX 4070 a mid-range card. Nvidia's 60-class cards used to be the one for the masses, with hits like GTX 1060 and GTX 1660 Ti, but fast-forward to today, and the RTX 4060 is all but a laughingstock.
While true mid-range performance now costs a bomb, relatively cheaper cards like the RTX 4060 are simply disappointing. Even AMD's competing offerings like the RX 7700 XT are far too expensive for what they offer, while still lagging behind Nvidia in key features. Buying a reasonably priced graphics card is not an option anymore if you want the best of performance and features without blowing a hole in your budget.
Intel Arc still can't be counted on
Intel's flagship product in its desktop GPU lineup, the Intel Arc A770 was a surprisingly solid entry by Team Blue in the gaming GPU duopoly of Nvidia and AMD. Even the Intel Arc A750 made sense for budget buyers looking for a sensibly priced graphics card. But, despite all the good things about Intel Arc, it's still a distant third option for gamers, with those targeting 1440p 100 FPS ray tracing performance not even having it in their rearview mirror.
Intel might revitalize the budget GPU market with its upcoming Battlemage cards, but I don't have high hopes. Intel's Arc architecture has had its fair share of driver issues and power consumption concerns. While this can be attributed to a first-gen architecture, I don't think Intel's second showing would be able to rectify everything that went wrong with Alchemist.
Nvidia and AMD might be all but absent from the budget segment next year, but people will still have the option of choosing older-generation GeForce or Radeon cards instead of similarly priced Intel Arc cards. For a truly three-way battle in the desktop GPU market, Intel will need another few years to perfect its architecture and come up with real flagship products. Until then, all we can do is lament about the AI obsession of Nvidia and AMD.
Will AI render your GPU upgrade dream artificial?
I'm the last person to sound the death knell for a longstanding technology for no reason. Gaming graphics cards have been, for positive and negative reasons, the highlight of PC gaming since forever. But, if GPU manufacturers continue their unrelenting focus on AI chips (which they logically would) and keep ignoring gamers (which they have been), it stands to reason that PC gaming will start to look very different very soon.
Affordability has already disappeared from the lexicon of the average gamer. Availability and performance might be next if something doesn't give. Let's hope the hype around AI stabilizes as the industry evolves, and GPU makers start thinking of gamers again, sooner rather than later.