With Computex 2026 coming up in Taiwan, GF Securities analysts reckon Nvidia will tackle the latest AI computing fashion head-on.
The current buzz is that CPUs matter more in agentic AI systems, which has led some to wonder whether GPUs lose clout inside AI clusters.
That chatter grew after Intel’s first-quarter earnings, when the market started eyeing a shift towards CPUs for agentic AI computing.
GF Securities thinks Nvidia will use Computex to show off its Vera CPUs and claim they beat x86 rivals from AMD and Intel. The analysts expect Nvidia to say Vera can deliver 1.5 times faster computing performance than competing x86 chips.
They said Nvidia is likely to make significant CPU announcements at the show, where Arm-based processors will be pitched as the new agentic AI workhorse. At the centre of the argument is Vera Rubin, Nvidia’s next platform for a market suddenly obsessed with orchestration, inference and autonomous software bots.
Nvidia is expected to focus on Vera Rubin CPUs, while Intel is tipped to talk up its entry-level Wildcat Lake processors. The two companies are heading into Computex with different headaches, although both will be trying to convince punters they have the right silicon for the next AI spendathon.
For Nvidia, the pitch is that its CPU story is not just some side quest while its GPUs keep printing money. GF Securities said inference workloads could change the CPU-to-GPU ratio inside AI clusters, which would help push Vera adoption.
The analysts expect Nvidia to claim Vera delivers 1.5 times faster performance, two times overall performance and four times higher density per rack than rival x86 chips from Intel and AMD.
That is a chunky boast, assuming the figures survive contact with real workloads rather than slideware theatre. They reckon Nvidia could ship 1.2 million CPU units in fiscal 2027 and 4.2 million in fiscal 2028.
Chipzilla, meanwhile, is expected to focus on Wildcat Lake, which GF Securities sees as a rival to the Fruity Cargo Cult Apple’s MacBook Neo laptops. The analysts expect agentic AI to expand the total addressable market for server CPUs, because every bot army needs something to boss the compute around.
They think agentic AI could eventually account for 30 per cent of inference computing. That would push the CPU market opportunity to $211 billion by 2030, assuming the AI boom does not faceplant before then.
GF Securities expects CPU demand to grow from 3.7 million units in 2026 to 16.3 million units in 2028.
