The usually reliable leaker MLID (Moores Law is Dead) sparked an interesting rumor over the weekend. He claims that Nvidia&s entry into the APU market will not only happen this year, but that the mobile combination of GPU and ARM CPU is already close to a concrete announcement.
Nvidia&s CEO Jensen himself confirmed at CES 2025 to journalists that Nvidia will soon introduce such a chip. MLID now claims that this chip was even supposed to be presented at CES – however, Nvidia reportedly decided against mentioning it in the keynote at the last minute. The reason for the last-minute cancellation is said to be that Jensen only wants to bring a "one hundred percent solution" to the market. This means that Nvidia&s approach must not only compete against Qualcomm but also against AMD&s Strix Halo. Conversely, this probably means that Nvidia must at least match the specifications of AMD&s flagship APU.
Furthermore, MLID claims to have learned that the CPU/GPU combination shown at 
Project DIGITS has a lot in common with the upcoming notebook SoC. However, it would have surprised us much more if Nvidia were to launch two completely different APU designs, where the synergies between the two versions are obvious.
Moreover, it is already considered certain that Nvidia wants to deliver the first Project Digits computers as early as May. This also means that the use in notebooks will be more a question of drivers and hardware crippling, and the chip itself is probably already practically finished.
However, it is still open whether Nvidia will sell the chips itself in the Windows market. After all, MediaTek is responsible for the ARM CPU part, under whose name the chip could also appear. Nvidia would then "only" receive license income, which in turn would not negatively affect the overall profit margin of the company. But these are just minor details, because for a few hours now, other rumor sources have also been churning out information on the topic:
Two Windows on Arm (WoA) chips are said to be presented at Computex 2025 in May: a high-end N1X model and a mid-range N1 chip. Since it is already assumed that Project DIGITS has a 512-bit memory bus, one could expect with a little imagination that the top model N1X could also have this feature. And the mid-range around the N1 chip could then compete with AMD&s Strix Halo and Apple&s M4 Pro thanks to 256 bits. Qualcomm would then only have the entry-level class for itself – with a 128-bit interface (at approx. 140 GB/s).
However, Nvidia might not reach so far out and the data buses of the N1/X will only be 128/256 bits wide. After all, it is also still necessary to sell its own high-end dedicated laptop GPUs – which have just been launched as the RTX 50 series.
Under Blackmagic Resolve, such APUs show themselves from their best side. Both Qualcomm and Apple already show today that mobile editing on a laptop can feel highly performant. This should be similar with Nvidia&s APU, especially since Nvidia can now finally also decode 4:2:2/10-bit in hardware.
;)