Nvidia wages chip war on two new fronts. But can it hold the line on all of them?

For a company that built its fortune selling graphics cards to gamers, Nvidia Corp (NASDAQ:NVDA, XETRA:NVD) has come a remarkably long way.

It now dominates AI data centres with a stranglehold that has made chief executive Jensen Huang one of the most powerful figures in technology.

But Nvidia appears unwilling to stop there.

This week, at its GTC Taipei event, the company unveiled RTX Spark, a superchip combining a Blackwell GPU with a 20-core Arm-based Grace CPU, up to 128 gigabytes of unified memory, and 300 gigabytes per second of memory bandwidth, all packaged into a thin Windows laptop.

The bigger question is whether this can do for AI PCs what discrete GeForce GPUs did for gaming laptops: Create a performance category that users can understand and that software developers can reliably target.

That is a genuinely hard problem.

The full RTX and DLSS 4.5 graphics stack is present from day one, with native anti-cheat support built in, which addresses one of the thorniest obstacles for Arm-based Windows gaming.

But Nvidia is simultaneously opening a second, arguably more consequential front.

Its new 88-core Vera data centre CPU claims a 50% performance gain over standard processors, fuelled by a 1.5x improvement in instructions per cycle from its custom Olympus cores, with a rack-scale architecture squeezing 256…