Nvidia owns the AI story, so why is AMD beating it on returns?

At the start of 2026, most investors would have confidently picked Nvidia stock as most likely to lead the market.

Jensen Huang still owns the AI stage, Nvidia NASDAQ:NVDA remains the default name in data-center spend, and the company has been expanding its CPU ambitions with Meta while touting a $1 trillion AI-chip opportunity through 2027.

However, the market has made a quieter call, with AMD doubling from its January 2 close of $227.15 to $455.19, while Nvidia has risen about 14% from $188.85 to $215.20.

That is a gap of roughly 86 percentage points.

Why the AI trade is quietly rotating

The first phase of AI was about training models, and Nvidia dominated that market.

The second phase is about inference, running those models in the real world and the hardware mix is changing with it.

The companies are moving from training to deployment, a shift that is lifting demand for server CPUs as much as GPUs.

AMD’s own latest results backed that up as Lisa Su said data center is now the “primary driver of our revenue and earnings growth,” adding that “inferencing and agentic AI” are increasing demand for high-performance CPUs and accelerators.

AMD’s first-quarter revenue was $10.3 billion, up 38% year on year, while data-center revenue jumped 57% to $5.8 billion.

That is the key re-rating as AMD is no longer being valued only as a GPU challenger, but as a CPU beneficiary in a market that is expanding faster than investors expected.

On its Q1 call, AMD said the server CPU addressable market could grow more than 35% a year to over $120 billion by 2030, up from a prior 18% view.

The company’s shares have outperformed Nvidia’s this year because of rising CPU demand and new deals.

AMD’s CPU-GPU stack is the surprise

The market is also starting to price in a strategic advantage that looks simple but matters a lot: AMD sells both CPUs and GPUs.

Ben Bajarin, CEO and principal analyst at Creative Strategies, told Reuters, “We like the integrated approach that AMD will bring with both CPUs and GPUs.”

That matters because AI systems are not one-chip machines anymore.

When a customer wants a tuned inference system or a more balanced training setup, a supplier that can design both sides of the workload has more room to optimize.

Wedbush’s Matt Bryson put the same shift in a line that sums up the mood on the Street: “The CPU is dead, long live the CPU.”

In simple words, the chip market is rediscovering the part of the data center it once treated as boring.

That is useful for AMD because it gives the company a second engine if AI GPU demand slows or arrives in waves.

Should investors switch from Nvidia to AMD?

Nvidia still owns the training ecosystem and the highest-quality AI franchise.

It is also pushing into CPUs themselves as Huang is increasingly talking up CPUs as AI moves into deployment, and that Nvidia’s CPUs are being positioned for agentic workloads and backend tasks such as database work.

But AMD has the cleaner near-term story right now because the CPU rotation is already showing up in revenue, margins, and share price.

Nvidia’s CPU plans are gaining attention, but AMD is already benefiting from the CPU boom today.