AMD Is Priced Like the No. 1 AI Chipmaker. It's Still No. 2

AMD has gained about 137% this year, nearly 150% in four months, and sits within 15% of its record high. The reason isn't hard to find: in its most recent quarter, the data-center business that used to be Nvidia's exclusive turf roughly doubled. This is a genuinely strong company story. But there is a second story buried inside the first one, and it is the one worth paying attention to before you add AMD to your list — the stock now trades as if AMD were the co-leader of AI chips, when the numbers say it is still a distant No. 2.

That gap between the rally and the business is the whole game. The easy money from AMD's re-rating has, in large part, already been made. What happens next depends on whether the business can grow into the price — which is a much harder question than whether AMD's data-center revenue is growing.

A real acceleration, on paper

Start with the numbers, because they are genuinely good. In the second quarter of 2026 AMD posted revenue of $11.5 billion, up 50% from a year earlier, with GAAP earnings of $1.38 a share and $1.66 on an adjusted basis. The engine was the data-center segment: $6.7 billion, up 107% year over year, and now 58% of all company revenue. AMD guided third-quarter revenue to roughly $13 billion, up about 41% year over year and a bit above what analysts had penciled in.

The forward story is real too. AMD's first rack-scale AI system, Helios, starts shipping this half of the year, and it has landed big-name customers — Microsoft, Meta, OpenAI and Oracle — while Anthropic has committed to deploying up to two gigawatts of AMD's MI450 chips in those racks beginning in early 2027. That is a credible bid to become the sanctioned "second source" to Nvidia that hyperscalers have been openly looking for. AInvest's aggregate signal still labels AMD a Buy. On the business itself, the market's enthusiasm is not fantasy.

The tell that the easy part is over

Here is what separates the two stories. After AMD beat estimates and raised guidance in early August, the stock fell more than 10% in extended trading. Read that again: a record quarter, a beat on both lines, and shares sold off hard. Then the stock climbed back toward its highs anyway.

That reaction is a useful signal. When a company beats-and-raises and still gets sold, it usually means the price had already moved ahead of the earnings. The re-rating — the part of the return that comes from investors simply deciding to pay more for each dollar of earnings — has substantially happened. The remaining gains have to be earned by the underlying numbers catching up, which is slower and less forgiving than a multiple re-rating.

The expensive second fiddle

Put AMD next to the actual leader and the valuation looks uncomfortable. Nvidia controls roughly 81% of the AI data-center chip market, versus AMD's roughly 10% — up from about 5% a couple of years ago, so the progress is real, but the gap is an order of magnitude. In Nvidia's last fiscal year, data-center revenue was about $194 billion; AMD's entire data-center segment was roughly $16.6 billion. Nvidia does that at a roughly 71% gross margin. AMD's gross margin is in the low 50s, and its full-company operating margin is around 11-17% depending on the quarter.

Yet it is AMD that carries the richer multiple — around 20 times sales and more than 100 times trailing earnings, versus Nvidia at about 17 times sales and roughly a quarter of that earnings multiple. For a growth investor, that is the central tension: you are paying a premium to the most profitable, dominant chip company in the world for the privilege of owning its much smaller, lower-margin challenger. The price is not crazy if AMD keeps compounding data-center revenue at triple digits. But that is a big "if," and the multiple leaves little room for merely good results.

That is the bear fact that has to be answered, not waved off. The question was never whether AMD wins some AI share — it clearly is. The question is whether it can win enough share, at enough margin, quickly enough to make a triple-digit earnings multiple reasonable within a reasonable time. If growth merely moderates from "blowout" to "great," or if the margin gap with Nvidia persists, that multiple can compress hard, and a stock that just doubled can give a lot of it back.

What this means for your judgment

The honest read, in the end, is that the market is not obviously wrong here. The business is genuinely accelerating, AMD is finally a credible challenger, and the AI-compute wave lifting it is secular, not a story. This is not the setup where a beaten-down, misunderstood stock is priced for doom — that is not what AMD is anymore. So this is not a call to short it or to call a top. It is a warning about what you would actually be buying.

When you buy AMD today you are not buying a bargain and you are not front-running a re-rating; you are paying full price for the assumption that the MI450 and Helios ramp lands on schedule, at volume, and with improving margins. Your margin of safety is delivery, not a cheap price. That means the invalidation is specific: the combination of data-center growth fading back toward the company's overall growth, or an earnings multiple that keeps expanding without the earnings to fill it. The thousand-percent gains of the last decade are history; the question now is whether the next leg is built on earnings catching up to a price that assumes they already have.