Wherever compute flows, Lisa Su invites those customers onto the stage.
This isn’t metaphorical—in AMD’s most recent earnings call, Lisa Su stood not beside chip engineers, but alongside representatives from major cloud providers and data center customers. That alone is a signal: AMD’s data center business has evolved from “selling chips” to “selling ecosystems.”
$70 Billion—It’s More Than Just a Number
AMD’s market cap has recently surged past $700 billion—a figure nearly unimaginable a decade ago, when AMD was still locked in fierce competition with Intel over PC and server CPU markets, and its stock price languished in single digits.
Today’s story is entirely different.
The core driver behind AMD’s rising market cap isn’t how many Ryzen processors it ships—or how popular Radeon graphics cards are among gamers. It’s the data center business.
Amid an explosion in AI compute demand, data centers are transforming from “enterprise IT infrastructure” into “the world’s most valuable real estate.” And AMD’s MI300 series accelerators landed precisely at this inflection point.
What AMD Got Right—Under NVIDIA’s “Shadow”
While everyone debates NVIDIA’s H100, B200, and GB200, AMD chose a path that may seem less glamorous—but far more pragmatic.
First: Don’t chase “best”—aim for “good enough.” NVIDIA’s flagship chips set the industry performance ceiling—but also the price ceiling. For many training and inference workloads, the MI300X delivers performance that’s already “good enough,” while its lower price directly boosts customer margins.
Second: Embrace openness. AMD’s ROCm software stack may lag CUDA in ecosystem breadth—but its open strategy is attracting customers eager to avoid vendor lock-in with NVIDIA. Especially for large-scale deployers, having a second viable option means greater negotiating power.
Third: Lisa Su’s personal IP. This isn’t a joke. Lisa Su’s reputation and approachability across the industry are genuine commercial assets. When clients negotiate with AMD, they’re speaking with a CEO who deeply understands both technology and real-world customer pain points—not a professional manager reciting quarterly numbers.
The Real Threat Isn’t in the GPU—It’s in the System
If AMD’s GPUs are still catching up to NVIDIA’s, its CPUs have nearly secured dominance in the data center.
EPYC server processors continue gaining penetration across cloud providers. Even more critically, AMD is integrating CPUs and GPUs into end-to-end system-level solutions—delivering everything from processors and accelerators to interconnect technologies, all under one roof.
What does that mean? Customers can engage with a single vendor and receive a complete, integrated data center compute solution. For procurement decision-makers, that’s vastly simpler than buying GPUs from NVIDIA, CPUs from Intel, and networking chips from Broadcom—and then assembling them themselves.
But After the Celebration Comes Caution
Behind that $700 billion market cap lies reality that demands sober assessment.
Though AMD’s data center revenue is growing rapidly, its base remains far smaller than NVIDIA’s. In fiscal year 2025, NVIDIA’s data center revenue exceeded $100 billion, while AMD’s hovered around $15 billion—representing not a difference of tens of percentage points, but an order-of-magnitude gap.
Meanwhile, AMD’s lofty valuation has already priced in several years of aggressive growth. If follow-on MI300-series products fail to sustain competitiveness—or if NVIDIA adjusts its pricing strategy—AMD’s stock could face significant volatility.
A More Intriguing Perspective
Yet viewed differently, AMD’s story may be even more compelling than NVIDIA’s.
NVIDIA’s success rests on the logic: “I have the best chip—so customers must come to me.” A classic supply-driven business model.
AMD’s success, by contrast, follows the logic: “Tell us what you need—we’ll deliver it.” It doesn’t chase industry-leading specs in any single metric; instead, it pursues optimal total cost-performance across full-stack solutions.
As AI compute demand shifts from “an arms race among a few tech giants” to “infrastructure for thousands of industries,” the latter logic may prove more durable.
Lisa Su never tried to become the next Jensen Huang. She chose to be herself—a pragmatic, open, customer-centric leader of a semiconductor company.
And the market, clearly, cast its vote—with a $70 billion endorsement.