ChatGPT could use 30,000 Nvidia GPUs in next upgrade

Market research firm TrendForce has reported that ChatGPT could benefit Nvidia by selling around 30,000 GPUs to OpenAI.

As it keeps growing and evolving, OpenAI’s ChatGPT is going to require even more power from hardware. The main source of this growth lies within graphics cards, something that Nvidia has been pushing since around 2017.

The next version of ChatGPT is expected to require around 30,000 GPUs, as the generative AI is ramping up to not only provide the current product, but also Microsoft Bing AI and other ambitious projects.

ChatGPT and Nvidia GPUs

However, don’t fret about it affecting GeForce products, like the RTX 4090 and 4080, as TrendForce reports that this will likely be based around the A100 GPU. This particular GPU is what currently powers Nvidia’s data centers. When implemented into a cluster, they will be able to soak in the various training data being implemented.

Nvidia A100 GPUs will cost a pretty penny

An example of what a deep learning workstation could cost can be found on UK retailer Scan’s website. The DGX Station outfitted with 28 Ampere GPUs costs nearly $200,000. However, the units that OpenAI will be using and investing in will presumably be placed into racks, rather than dedicated boxes.

It’s unknown if ChatGPT will be using next-gen tech like the upcoming Hopper architecture, which is next in line from Nvidia. We also do not know if it’ll be using what Nvidia dubs ‘Ampere Next Next’. With the way things are shaping up, it might be running on what’s available, rather than holding out for future products.

OpenAI’s ChatGPT has led to multiple successes in the last few weeks, with Microsoft furthering its multi-billion dollar investment and starting to implement AI into Bing and Office. Meanwhile, Google has begun to launch its own generative chat AI, Bard.

We previously reported that Chinese games publisher, NetEase, will begin to utilize ChatGPT to generate NPC dialogue and story outcomes depending on what the player input. Needless to say, it’s another boon for the hardware industry at large.