AMD Keeps Falling Below $500 After Winning — That's a Math Problem, Not a Broken Business

On August 4, 2026, AMD reported the strongest quarter of its modern life: record revenue, a beat on both the top and bottom lines, data-center sales that more than doubled. The stock closed below $500. In mid-September it slipped back under $500 again, to the mid-$490s, even as Wall Street keeps pushing price targets higher.

That pattern looks like a contradiction — a company that keeps winning, a stock that keeps failing. It isn't. It's a company compounding revenue at roughly 50% a year while trading at about 60 times earnings. Which is another way of saying the price already assumes AMD is perfect. Numbers only confirm that; they almost never improve it.

The quarter was exceptional. The result wasn't.

Here is what AMD actually printed for the June quarter.

Revenue came in at $11.54 billion, up 50% from $7.69 billion a year earlier. Adjusted earnings were $1.66 a share, edging past the $1.62 analysts expected. The engine was the data center: that segment's revenue hit $6.7 billion, up 107%, and now makes up 58% of AMD's total revenue, up from 42% a year ago. For the third quarter the company guided to roughly $13 billion in revenue, growth of about 41%.

That is a genuinely strong report, and the market treated it as a disappointment. The stock fell about 5% to $493.41 on the print, then dropped another 7% to $482.05 the next session. It had crossed below $500.

The street's reaction was boiled down cleanly by Futurum's Shay Boloor: the stock "was priced for an exceptional result, and this was not an exceptional result". At nearly 60 times earnings, AMD is valued for a blowout, not a beat. Two small details in the report fed the disappointment. Capital expenditures jumped to $808 million from $282 million a year earlier — investors read higher spending as a cost, not a growth signal. And management warned of a softer PC market in the second half, a side business that many buyers had already stopped caring about.

The sharper reminder of AMD's actual position came the next day. Elon Musk announced that SpaceX would buy its AI chips exclusively from Nvidia, citing Nvidia's Vera Rubin architecture as superior. A marquee, named customer publicly choosing the rival is the clearest possible statement of AMD's role in this market: it is the credible number two, not the leader. The hyperscalers that do buy from AMD — OpenAI and Meta together committed to 6 gigawatts each of AMD accelerator capacity, with Azure and Oracle early customers of its rack systems — are buying a second source and negotiating leverage, not a replacement for Nvidia.

Business quality and stock price are different things

The useful lesson here is not that AMD is overvalued or undervalued. It's that the two questions — is the business growing, and is the price reasonable — have no automatic relationship.

AMD's underlying data is getting better, not worse. Its newest AI accelerator family, the MI400 series, features a flagship GPU with 432 gigabytes of HBM4 memory, and its Helios rack — 72 accelerators in a single unit priced around $5.25 million — was in full production by August, with first customer shipments scheduled for late in the third quarter. The company is betting on inference, a market it projects growing about 80% a year, where it can win on memory and bandwidth rather than raw training throughput. Analysts responded by raising targets: one reset its price target to $641 on August 25, implying roughly 40% upside from the share price at the time. Over the past year the stock has still more than doubled even after the selloffs.

What's holding AMD back isn't the trajectory of its data center business. It's the multiple. When you pay 60 times earnings, every quarter that confirms rather than exceeds resets the bar even higher, and every competitive headline — one customer to Nvidia, a capex line that surprises — gets amplified. The thing that would actually change this story is visible in the data: data-center growth slowing toward the guided rate, another marquee customer defecting to Nvidia, or margins compressed by an expensive ramp. Watch those, not the daily headlines.

Until then, the recurring trip below $500 is the market re-testing whether an exceptional business deserves an exceptional price. AMD keeps delivering its half of that bargain. The price keeps enforcing its own. Your job is to decide which half you believe will move first — not to assume the falling share price means the company is falling with it.