OpenAI could ditch Nvidia GPUs, handing a lifeline to PC gamers desperately seeking affordable cards

Well, this wasn’t in our top ten 2026 expectations. According to eight sources “familiar with the matter,” OpenAI is “unsatisfied with some of Nvidia’s latest artificial intelligence chips”, and actively seeking to move away from Nvidia GPUs in favor of more specialized processors. Were this to cause a significant reduction in demand for Nvidia GPUs for AI, it could mean GPUs for graphics cards becoming more readily available, though it remains far too early to tell exactly how this situation will play out.

The reason for this potential shift in demand is that while Nvidia GPUs remain dominant for training AI models, the power of these chips for providing the front-end responses to customers – known as inference – is being challenged, with application-specific integrated circuits (ASICs) being increasingly favored for these tasks over general-purpose GPUs.

The actual impact of this potential drop in demand would certainly be significant, but not quite as huge as some PC gamers might hope. According to Reuters, OpenAI is looking to shift about 10% of its hardware to these new inference-specific products, seemingly leaving 90% of its computing still dependent on GPUs from both AMD and Nvidia.

It’s also unclear if this will directly help the situation we’re currently facing with inflated RAM and SSD prices. A chunk of this demand and subsequent price increase comes from VRAM demand for GPUs used in AI data centers, but those data centers will likely still need conventional RAM and SSD storage, even if they rely less on GPUs.

Still, if there is a larger overall shift in the AI market towards a different type of processor and data center setup, there is definitely still potential for the very direct competition between gamers (wanting GPUs for PCs) and AI companies (wanting them for data centers) to be relaxed.

As to who is making these new ASICs, Google has already largely shifted to using these types of processors, using its in-house-developed chips. Meanwhile, companies such as Cerebras and Groq are also seeing heavy investment from Nvidia and OpenAI. These chips still need to be produced by chip manufacturers, such as TSMC, so there’s still a strain on overall chip production, but we can live in hope that the ballooning prices for high-end Nvidia cards might drop a little sooner than we previously thought.