Nvidia vs Intel: Jensen Huang braces investors for renewed battle as chip wars reignite

Nvidia CEO Jensen Huang is aggressively preparing investors for a new phase of competition against long time rivals Intel and AMD. While Nvidia remains a central player in AI training, Huang is increasingly emphasizing the importance of the general-purpose CPU. He noted that while computing was once 90% CPU-based and 10% accelerated, the ratio had reversed in recent years. However, the CPU is making a comeback as an essential component for agentic AI, as companies shift from building models to deploying them.

CPUs vs. GPUs: Analyzing the shift in AI infrastructure

CPUs and GPUs have had distinct computing tasks for decades. CPUs are general purpose chips designed to handle complex tasks at reasonable speed. On the other hand, GPUs specialize in carrying out simpler calculations in parallel, performing thousands of operations simultaneously.

In this connection, Ben Barjarin, an analyst at Creative Strategies said: “AI companies are increasingly deploying agents that can independently carry out tasks such as writing code, shifting through documents and research reports."

Nvidia emerges as central CPU player to challenge Intel and AMD

Nvidia officially announced a deal with Meta Platforms, marking a significant strategic shift. Unlike Nvidia’s current AI servers where each CPU is paired with multiple GPUs, this deal focuses on the large-scale deployment of standalone Grace and Vera CPUs.

Huang’s view is that the architecture is designed for high-performance data processing. The primary reason is that most modern computing problems are data-driven with artificial intelligence being a prime example. 

In this regard, Dave Altavilla, principal analyst at HotTech Vision and Analysis said: “Nvidia is aiming to prove that the CPT type once supplied primarily by Intel “is no longer assumed the default foundation of modern computer infrastructure.