Jensen Huang says GPUs will not replace CPUs in AI computing
TechRadar highlights comments from Nvidia co-founder Jensen Huang, who says GPUs will not replace CPUs as AI computing gives both processor types important roles.

Key takeaways
- Jensen Huang says GPUs will not remove the need for CPUs.
- TechRadar says Huang made the comments in a televised PBS interview.
- TechRadar says CPUs can offer lower latency and quick access to memory for inference.
- TechRadar says Nvidia later became the most valuable company globally.
In this article
What Jensen Huang said
According to TechRadar, Jensen Huang has made a clear point. TechRadar published the item on 20 November 2026.
He says GPUs will not remove the need for CPUs. GPUs are graphics processing units, and CPUs are central processing units. Huang is a co-founder and former CEO of Nvidia.
He made the remarks in a televised PBS interview. The interview took place when the financial crisis was at its height. TechRadar says Huang used the appearance to challenge the idea that GPUs were mainly for gaming.
He argued that these chips were part of the future of computing. He did not present them as a niche product for games. The article is part of the TechRadar Pro QOTD project. It presents the remarks as a way to understand a major technology figure. The excerpt from TechRadar says the CPU has become important again.
Why CPUs still matter
CPUs have been the base of computing for decades. Huang stressed that a computer cannot work without one. That view places the CPU at the centre of the PC. It also matters as graphics chips gain attention.
He explained that GPUs are strong at parallel processing. This means many calculations can run at the same time. In his demonstration, programmers used Nvidia hardware to complete tasks faster than with conventional chips.
Even so, he said both types of processor have a place in modern machines. The excerpt from TechRadar makes a similar point. It says GPUs have gained attention over the last decade. It also says conventional processors matter again.
The point from Huang was not that GPUs are unimportant. It was that they do not replace the processor at the centre of the computer.
A difficult period for Nvidia
TechRadar notes that Nvidia was struggling during that period. It also notes that Huang departed Nvidia on controversial terms. The company still spent heavily on research and development. That choice helped it move through a difficult phase.
The source says Nvidia later became the most valuable company globally. It says this happened in recent years. This background helps explain why the comments from Huang drew attention.
They came from a leader whose company helped shape the GPU market. He defended the role of CPUs at the same time. TechRadar says conventional computing processors have been the foundation of computing for decades.
AI changes the balance

GPUs remain in high demand because they perform well in AI data centres. TechRadar says demand for GPUs has surged. These systems need many chips to train and run models.
However, the source says the industry is now turning to CPUs for inference. Inference is the stage where a trained model responds to a request. CPUs can offer lower latency, more flexibility and quick access to memory.
Those qualities make them useful for large language model work. In that work, responses need to be fast and adaptable. TechRadar says a large share of computing resources now goes to deployment as well as training. That means chips are needed after a model is built.
TechRadar says this mix is becoming more important. AI computing is not limited to building models. It also includes serving them to users. That gives CPUs a larger share of the work.
What it means for computing
The remarks suggest that the AI era does not make CPUs obsolete. Instead, it gives them a new role. Training and graphics intensive tasks may still lean on GPUs. Inference, general computing and low latency work may rely more on CPUs.
For readers, the main point is balance. Modern computers are likely to use both types of processor. Each can handle the jobs it does best. TechRadar presents the comments from Huang as part of the QOTD project. That project looks at ideas from major technology figures.
The source does not provide benchmark results or specific product advice. It focuses on the strategic point that GPUs and CPUs can coexist. That message is relevant as AI systems move from training to wider deployment.
Sources
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



