Lisa Su just telegraphed the plan. And it's bigger than most people realize.
Two years ago, $AMD was the alternative. The 'good enough' chip. The one you bought when $NVDA was sold out or too expensive. Today, that narrative isn't just dying β it's getting buried under $10.27 billion in quarterly revenue and a data center business that's growing faster than almost anything in Silicon Valley.
The MI400 is coming. CDNA 4 architecture. Rumored 2026 launch. And it's the kind of chip that could reshape who owns the AI training stack for the next five years.
The Numbers Tell the Real Story
Let's cut through the noise. Here's where $AMD actually sits right now:
| Metric | Value |
|---|---|
| Market Cap | $839.5B |
| Revenue (TTM) | $37.45B (+37.8%) |
| Net Income | $4.93B |
| Gross Margin | 53.1% |
| Forward P/E | 39.1x |
| FCF | $7.17B |
| Current Price | ~$515 |
Quarterly revenue has gone parabolic: $7.69B in Q2 '25, then $9.25B, then $10.27B, holding at $10.25B in Q1 '26. That's not a fluke. That's a structural shift.
The MI300X alone is running at over $5 billion in annualized data center revenue. Think about that. One chip family generating more revenue than entire semiconductor companies.
The MI400 Thesis: Why This Changes Everything
Here's the part the bears are ignoring. The MI400 isn't just an incremental refresh. It's CDNA 4 β a complete architectural rethink built for the era when AI training clusters are measured in tens of thousands of GPUs.
The addressable market for AI accelerators is projected to hit $400 billion+ by 2027. That's not a niche. That's a generational opportunity. And right now, $NVDA owns roughly 80% of it. The question isn't whether there's room for a second player. The question is whether $AMD can actually take share.
The answer, based on every data point we have, is yes.
Competitive Positioning: The Underrated Moat
Nvidia charges premium pricing on everything. That works when you're the only game in town. But hyperscalers β Microsoft, Meta, Google, Amazon β are actively diversifying their silicon supply chains. They don't want one vendor holding the keys.
AMD is positioning the MI400 at lower price points with competitive performance. That's not a race to the bottom. That's a share-gain playbook. Every dollar saved on compute is a dollar hyperscalers can redeploy into more capacity.
The trailing P/E of 171 looks terrifying until you realize it's inflated by acquisition amortization. The forward P/E of 39.1x on 37.8% revenue growth? That's actually reasonable for a company in this growth phase.
The Bottom Line
$AMD trades at 22.4x sales with a beta of 2.49. This stock moves fast. It punishes the weak-handed and rewards conviction.
Data center is now the majority of AMD's revenue. The MI300X proved they can ship at scale. The MI400 is about to prove they can compete at the frontier. The TAM is massive, the margins are expanding (53.1% gross), and free cash flow sits at $7.17 billion.
AMD isn't the underdog anymore. They're the credible second source in the most important technology cycle of the decade. If the MI400 delivers even 80% of what the roadmap suggests, this stock has fundamentally different math than the market is pricing today.
The comeback isn't coming. It's here.
β The Signal
