Why Nvidia Still Looks Like the Trade, Despite the Noise
Custom silicon is eating share, AMD is getting credible, and chip stocks have been volatile all month. The fundamentals underneath Nvidia are still doing most of the talking.
AI chip stocks have had a rough few weeks of headlines — sharp single-day drops, recoveries, more drops. Strip the noise out and the underlying numbers are still the story.
The numbers that matter
Nvidia closed fiscal 2026 with $215.9 billion in revenue, up 65% year over year. The data center segment alone did $62.3 billion in the fourth quarter, up 75% from a year earlier. Jensen Huang's framing — that AI compute is becoming core global infrastructure, not a cyclical buildout — is easier to defend with growth rates like that sitting underneath it.
$215.9B
FY26 revenue
+65%
YoY growth
$62.3B
Q4 data center
25.4x
Forward P/E
A forward P/E of 25.4 on that growth rate is, on a pure growth-adjusted basis, not an aggressive multiple for a company still putting up triple-digit data-center growth. That's the bull case in one sentence.
The risk underneath the bull case
The thing to actually watch isn't a single bad quarter — it's market structure. Custom silicon (the chips hyperscalers design in-house, think Google's TPUs or Amazon's Trainium) made up about 20.9% of the AI chip market in 2025 and is projected to reach 27.8% in 2026. That's a meaningful and fast-moving share shift, and it's coming from Nvidia's biggest customers, not from a scrappy outsider.
AMD is the more traditional threat, and it's becoming a more credible one — its pitch is competitive performance at a noticeably lower price point, which matters more as AI infrastructure spending shifts from "land grab" to "cost discipline."
- Custom ASICs (TPU, Trainium, and similar) are the structural risk to watch, not quarterly earnings beats.
- AMD's price-performance pitch gets more attractive as capex budgets tighten.
- Data-center growth, not gaming or auto, is still the number that decides the stock.
None of this breaks the thesis. It does mean the next leg of the Nvidia story is less about whether AI compute demand is real — that question is settled — and more about how much of the next dollar of that demand Nvidia gets to keep.
This is market commentary, not investment advice — markets are volatile and past performance doesn't predict future results. If you're making portfolio decisions, it's worth talking to a licensed financial advisor who knows your full picture.
Keep reading
AI Daily Brief: The Mythos 5 Export Saga, OpenAI Buys the Python Tooling Stack
An export-control fight over Anthropic's Mythos 5 is escalating fast, OpenAI just bought the team behind uv and ruff, and Gemini 3.5 Pro is still stuck in preview. Here's what actually matters today.
Agentic AI Just Crossed Its Tipping Point in the Enterprise
Adoption numbers are up and to the right, but there's a wide gap between companies that have agents and companies that have agents doing real work. That gap is this year's whole story.