NEW YORK — Advanced Micro Devices spent Tuesday absorbing pressure it did not generate. The stock fell $2.57 to $176.38 — a 1.4% decline — as investors across the chip sector moved to the sidelines ahead of NVIDIA’s fiscal second-quarter earnings Wednesday evening. AMD’s own AI business is growing. The stock’s problem, for now, is that it grows in NVIDIA’s shadow.
The MI325X, AMD’s flagship AI training accelerator, shipped to hyperscale customers ahead of the company’s original schedule according to multiple supply-chain sources cited in analyst notes through August. That timeline advantage matters because the chip’s successor — the MI350X, built on TSMC’s 3nm process — is expected to enter volume production in the fourth quarter of this year. AMD’s roadmap credibility is higher now than it was eighteen months ago, when the MI300X launched to skepticism about whether it could actually challenge NVIDIA’s H100 at the data-center level.
The answer, measured by revenue, is that it has — partially. AMD’s data center segment, which encompasses AI accelerators, server processors, and related products, generated $4.9 billion in revenue during the second quarter of 2026, up 121% year-over-year. That number looks enormous in isolation. In context, NVIDIA’s data center revenue for the equivalent period is expected to come in near $87 billion. The market AMD has captured is real but represents somewhere around 5% of the category NVIDIA currently dominates.
CEO Lisa Su said in the most recent earnings call that the company expected full-year 2026 AI accelerator revenue to exceed $10 billion, up from a prior target of $8 billion. That revision reflected faster-than-anticipated customer pull-through on the MI325X and early purchase commitments on the MI350X. The Wall Street consensus has moved accordingly: most sell-side models now estimate AMD’s AI accelerator revenue at $10.5 billion to $11 billion for the full year.
The PC CPU business, where AMD has spent the better part of a decade making genuine market share gains against Intel, produced mixed results. AMD’s client segment revenue was $1.8 billion in Q2, up 49% from a year earlier, driven by the Ryzen AI 300 series — the first consumer laptop chips AMD has shipped with neural processing units capable of meeting Microsoft’s Copilot+ specification. Intel is trying to close the same gap with its Lunar Lake and Arrow Lake processors, but AMD entered the AI-PC cycle with a measurable lead in silicon shipping to OEMs.
The server processor business, sold under the EPYC brand, is the most structurally compelling part of AMD’s portfolio from a competitive standpoint. EPYC now holds an estimated 34% share of x86 server processor units, according to Mercury Research data cited by multiple analysts. Three years ago, that number was under 10%. Intel’s response — the Xeon 6 family — has received positive technical reviews but has not stopped AMD’s share gains in cloud provider deployments. Microsoft Azure, Google Cloud, and Oracle Cloud have all expanded their AMD-based instance families in 2026.
NVIDIA’s earnings after Wednesday’s close are the most important single event for AMD’s stock in the near term, even though AMD is not reporting. If NVIDIA’s results reveal any demand softness, customer concentration risk, or supply constraints that haven’t been priced in, AMD’s relative positioning improves in perception — even if the absolute market sizes don’t change immediately. Conversely, a blowout NVIDIA quarter reinforces the narrative that AMD is a distant second in a category where second place matters much less than in traditional processor markets.
At $176.38, AMD trades at approximately 43 times forward earnings estimates for calendar year 2026, according to FactSet consensus data. That multiple reflects the company’s AI upside optionality more than its current earnings power. The PC and server CPU businesses are growing steadily but not spectacularly. The AI accelerator business is growing spectacularly but is priced for continued growth. Any sign of execution stumble on the MI350X timeline or customer adoption would compress the multiple quickly.
The July earnings call contained one detail worth watching: Su declined to give specific MI325X customer names, citing competitive sensitivity. That reticence is standard for the industry, but it has left analysts estimating adoption curves from component shipment data and OEM purchase patterns rather than direct disclosure. The uncertainty around actual installed base — versus chips shipped — is the largest unresolved question in AMD’s AI story heading into the second half of 2026.
What the market is watching: NVIDIA results Wednesday, which will function as a read-through for the entire AI semiconductor ecosystem. Beyond that, AMD’s own next earnings call, tentatively scheduled for late October, will need to show MI350X momentum to sustain the current valuation. Any slip in the AI accelerator revenue trajectory toward $9 billion or below would test the stock’s $170 support level.
Related Posts
