NVIDIA will now release new GPU architecture every year, 'Rubin' is teed-up for 2025

On average NVIDIA has released new GPU architecture every two years, but due to the overwhelming demand for new chips to power AI models the company is accelerating its manufacturing schedule and therefore the time between generations.

2

NVIDIA recently reported its first-quarter earnings where the company revealed it made an astounding $14 billion in profit from its dedicated AI chips, which have been in extremely high demand since the AI boom. Companies such as Meta, Amazon, Google, and OpenAI want NVIDIA's advanced GPUs to train and power their new artificial intelligence models. NVIDIA plans on meeting this overwhelming demand by increasing the frequency it releases new GPU architecture generations.

Notably, NVIDIA released new architecture approximately every two years, with Ampere releasing in 2020, Hopper in 2022, and Blackwell in 2024 (rumored to release at end of the year). But NVIDIA's CEO Jensen Huang announced, "after Blackwell, there's another chip. We're on a one-year rhythm." Analyst Ming-Chi Kuo reported earlier in the month that NVIDIA was planning on releasing its next-generation GPU architecture codenamed "Rubin" in 2025, which would likely produce an R100 AI GPU that will be released sometime next year.

It appears Kuo was spot on with that prediction. Huang was questioned on the success of the coming Blackwell-based AI GPUs (B100), and if it will be adopted by companies that recently purchased large swaths of Hopper-based AI GPUs (H100, H200). Huang responded by saying all Blackwell AI GPUs are backwards compatible and run the same software as Hopper-based AI GPUs, meaning buyers can "easily transition from H100 to H200 to B100" within their existing data centers.

Furthermore, Huang said that AI GPUs aren't the only chips that will be accelerated, with the NVIDIA CEO saying, "We're going to take them all forward at a very fast clip." Perhaps this means we will begin to see gaming-focussed GPUs every year.

"New CPUs, new GPUs, new networking NICs, new switches... a mountain of chips are coming," he says