AMD Grew Revenue 50% and Still Got Sold. Make That Make Sense.

Okay, so this one genuinely caught me off guard.

Advanced Micro Devices reported Q2 2026 earnings this week. Revenue came in at $11.54 billion — up 50% year over year. Adjusted gross margin hit 56%. Data center EBIT margin reached 31%. By basically any traditional measure, this was a great quarter. Maybe even a remarkable one.

And the stock dropped 7–8%.

I've been following earnings long enough to know that "sell the news" is a real thing. But a 50% revenue growth quarter getting punished like that still makes you stop and stare at the screen for a second. What is the market actually pricing in right now? And is this a buying opportunity or a warning sign? Those are two very different things, and I think the answer depends on which part of the AI trade you're looking at.

The Numbers Were Actually Good. Like, Really Good.

Let me just lay out what AMD actually reported, because I think the narrative around the selloff can obscure how strong the underlying business has gotten.

Fifty percent year-over-year revenue growth to $11.54 billion is not a number you shrug at. For context, AMD was doing roughly $5.8 billion in quarterly revenue at this point in 2024. The company has effectively doubled its top line in two years, mostly on the back of data center demand — which now accounts for 58% of total revenue.

That data center EBIT margin of 31% is also worth flagging. This isn't a company burning cash to chase growth. They're converting AI infrastructure spending into real operating profit. The 56% adjusted gross margin is up meaningfully from where AMD was sitting even 18 months ago, when the market was still skeptical they could compete with Nvidia in the GPU space.

Upcoming catalysts include Helios rack-scale commercialization, EFB packaging advances, and continued EPYC CPU adoption — which, if you've been following the server market, has been quietly taking market share from Intel in a way that doesn't get nearly enough coverage.

So what happened? Why did a quarter like this get treated like bad news?

Here's Where the AI Trade Gets Weird

Here's what I actually think is going on: the market isn't selling AMD because the results were bad. It's selling AMD because the results weren't transcendent. And that's a very different problem.

When a stock is priced for perfection — and a lot of AI names right now are priced somewhere between "perfection" and "the robots definitely handle everything by Q3" — a merely excellent quarter creates a gap between expectation and reality. It's not that 50% growth is disappointing. It's that some investors had already pulled forward several quarters of that growth into the price.

MarketWatch ran a piece today comparing the current AI excitement to 1999. And look, I know that comparison gets made constantly and is usually overdone. But some of the signals they're pointing to are hard to dismiss — the giddy confidence that this time the technology is so transformative that normal valuation guardrails don't apply. That's the same energy from 25 years ago, just with different ticker symbols.

And I'll be honest — this is the part that genuinely worries me. Not AMD specifically. AMD's fundamentals are real. But the environment around AI stocks right now has a circular quality to it where price movement itself becomes the justification for further buying, and any company that can credibly claim AI relevance gets a valuation premium that may or may not be warranted.

The Other Side of the Ledger: Cerebras Just Fell Off a Cliff

The AMD selloff gets even more interesting when you put it next to what happened to Cerebras Systems (CBRS) this week.

Cerebras is one of those companies that has become almost a symbol of the AI trade — a high-profile AI chip designer that went public with enormous fanfare and has been treated by parts of the market as a pure-play bet on silicon intelligence. This week, they reported earnings. And the stock plunged.

I don't have the full Cerebras breakdown in front of me, but the pattern here is telling. You've got AMD — a real, profitable, 50%-growth business — selling off. And you've got Cerebras — a much earlier-stage story with a valuation that required near-infinite optimism — also selling off, but harder. When both the "proven grower" and the "pure story stock" are getting hit simultaneously, that's usually a sign the market is doing some repricing across the whole sector, not just reacting to individual reports.

Going a step further — the neocloud space is showing the same bifurcation. Nebius just reported AI cloud revenue up 454% year over year. Which sounds insane. And it is impressive. But analysis out today flags that Nebius trades at a premium valuation that becomes harder to justify as capex rises and debt accumulates. CoreWeave, by contrast, is getting called out as the cheaper option with better cash flow dynamics and stronger scale. The market is starting to care about unit economics again, not just growth rates. That's a change worth paying attention to.

AI Is Everywhere — Including the Fed's Headache File

Here's the part of the AI story that doesn't get talked about enough in the context of stocks: the infrastructure buildout is itself an inflation input.

CNBC had a good piece on this. Tech leaders keep promising that AI will be disinflationary — Sam Altman's "intelligence too cheap to meter" line, Musk talking about extreme abundance, Masayoshi Son predicting a 40% drop in prices. These are real arguments. The productivity gains from AI, if they materialize at scale, could genuinely help bring costs down over time.

But right now, we're in the build phase. Data centers need land, power, cooling, construction workers, specialized components, and enormous amounts of capital. All of that is inflationary in the near term. The Fed is watching a situation where the same technology everyone says will solve inflation is currently contributing to it, through a massive investment boom that's pushing up wages and energy demand in specific markets.

And because AI is increasingly moving currency markets too — companies are executing large cross-border capital flows to fund data center expansion globally — the effects are rippling into places economists don't usually think to look. The traditional inflation model just doesn't fully account for a build-out of this scale and speed.

Meanwhile, markets were rising Thursday morning ahead of yet another inflation reading — the CPI print that Wall Street has been obsessively tracking all year. Expectations going in were modest: 0.1% monthly on the headline, 0.2% on core, with annual readings of 3.4% and 2.5% respectively. The fact that futures were up on "rate hike bets" tells you something interesting — the market has gotten to a point where it's pricing in potential Fed action without panicking about it. That's either maturity or complacency. Not sure which yet.

So Is AMD Actually a Buy Here?

Look, I could be wrong here, but — yes, probably, if you have a medium-term horizon.

The 7–8% post-earnings drop on a 50% growth quarter looks a lot more like a valuation reset than a business deterioration story. The fundamentals are real: data center at 58% of revenue and growing, margins expanding, multiple product cycles ahead with Helios and EPYC. This isn't a company that missed on substance.

What it missed on is the implied expectation baked into a multiple that requires the next few years to be as good as the last few. If you believe data center spending stays elevated — and given the AI infrastructure arms race, there's a reasonable case for that — then AMD at a post-earnings discount is arguably the cleanest way to own that theme with a company that's already demonstrating real profitability.

The more speculative AI names? That's where I'd be more cautious. When Cerebras is plunging on earnings and Nebius is getting scrutinized for its debt trajectory, the market is telling you that the "any AI exposure is good AI exposure" era might be ending. We might be entering a phase where differentiation actually matters — where profitable, scaled businesses get rewarded and pre-revenue stories get repriced back toward earth.

That isn't a crash. It's actually healthy. It's what markets are supposed to do. The AI memory stocks story from earlier this year had a similar shape — great fundamentals, strange post-earnings behavior — and that sector eventually found its footing. And if you want to understand how earnings reports actually drive these moves, here's a breakdown of how to read one before the market reacts.

The 1999 comparison from MarketWatch is worth filing away, not as an immediate alarm, but as a posture check. In 1999, the companies that survived and thrived weren't the ones with the biggest stories — they were the ones with actual business models. Amazon survived. Pets.com didn't. That sorting process is what seems to be starting now in AI, three years into this boom.

AMD has a business model. A very good one, as it turns out.

What to Watch If You're Holding AI Stocks Right Now

• If you own AMD: The 7–8% selloff lands the stock at a more defensible valuation. Watch the Helios rack-scale launch timeline — if that commercializes on schedule in late 2026, the next earnings setup gets much cleaner. Data center EBIT margin above 30% is the threshold to hold.

• If you're eyeing Cerebras or other pure-play AI story stocks: Wait for at least one more quarter of data. Post-IPO euphoria is still unwinding, and "plunged on earnings" is a yellow flag worth respecting. Let the dust settle.

• If you're comparing CoreWeave vs. Nebius in the neocloud space: The market is increasingly rewarding cash flow discipline over clean balance sheets alone. CoreWeave's lower reliance on prepayment funding is a genuine differentiator here — Nebius's premium is vulnerable as capex rises.

• For index fund holders watching the AI concentration risk: The S&P 500's exposure to AI-adjacent names is significant. If the sector repricing accelerates, diversification into value or international names isn't a bad idea. The long-term capital gains tax math also matters here — if you're sitting on big AI gains, think hard before triggering a short-term event.

• On the CPI print: If Thursday's inflation data comes in hotter than the 3.4%/2.5% expectations, expect the rate-hike bet to get louder — and growth stocks, AI names included, to face another leg of pressure. That's the macro wildcard hanging over all of this right now.