Nvidia’s AI advantage is increasingly moving beyond its powerful graphics processing units (GPUs), as the company positions itself to control more of the infrastructure powering the next generation of artificial intelligence data centres.
For years, Nvidia’s dominance in AI was largely tied to its high-performance GPUs, which became essential as companies raced to build increasingly powerful AI systems. But competition has intensified, with major cloud companies such as Amazon and Google developing their own chips. While that has raised questions about the durability of Nvidia’s lead, the company is now betting that its biggest advantage may lie in the wider systems surrounding the GPU.
The shift is becoming more important as AI data centres grow to enormous scales and consume massive amounts of computing power. Nvidia’s new Vera Rubin architecture combines its Rubin GPU with other specialised components, including the Vera CPU, Groq 3 LPX inference accelerator, storage and networking systems. Rather than simply processing AI workloads, these components are designed to ensure that data moves efficiently to where it is needed, reducing bottlenecks and improving overall performance.
Nvidia says the Vera CPU can deliver up to a threefold improvement in certain data operations by helping manage the flow of information between storage and computing systems. The challenge is becoming increasingly critical as companies seek to reduce the energy required to generate AI tokens and squeeze more performance from every watt of power. Other AI companies are tackling the same problem differently. OpenAI, for example, has focused on reducing data movement through its Jalapeño chip, keeping more of the workload within a connected system to improve speed and efficiency.
The emerging competition therefore goes beyond who can build the fastest AI chip. As AI infrastructure becomes larger and more complex, controlling how processors, memory, storage and networks work together could become just as important as raw GPU performance. Nvidia will still face competition from chipmakers and hyperscalers, but its growing presence across the wider AI infrastructure gives it another potential advantage — and one that could prove increasingly valuable as the global AI race moves from simply adding computing power to making every unit of power work harder.
source: techcrunch
