AMD Is Up 160%. The Upgrades Are Just Catching Up

On August 4, AMD reported a quarter that beat Wall Street on both revenue and earnings, guided the next quarter above consensus, and watched its stock fall more than 10% in extended trading. An investor can search a long time for a cleaner demonstration that a good report and a falling price are not a contradiction.

Then, within weeks, the upgrades started landing again. CLSA pushed its target to $710 from $575. Raymond James moved AMD to Strong Buy at $641, up from $565. Piper Sandler opened coverage with a Buy and a $600 target. Same company, same numbers, opposite verdicts in the same stretch of weeks. The way to reconcile them is to understand what the analysts are really buying — and it is not the past quarter.

The quarter that got sold

The blunt explanation for the drop: AMD had nearly tripled over the prior year, and a beat only satisfies the expectation it meets. The numbers were good, but not good enough for a share price that had already moved. Q3 guidance of roughly $13 billion came in above the street's $12.5 billion consensus but below the $14 billion the most optimistic buyers had been hoping for. After a 160% run in a calendar year, the marginal buyer's bar is not "better than last year" but "better than the fantasy." That is a high bar no matter how strong the underlying business is.

The magnifier is valuation. AMD trades at roughly 74 times forward adjusted earnings, against a semiconductor industry average near 25 times. The market is paying a premium that assumes the growth story continues at speed. Good news meets that premium and, at the margin, gets sold.

The thing the upgrades are chasing

The upgrades are not decoration. They point at a genuine structural change — arguably the most important thing happening at AMD. The company is being re-rated from a PC and server CPU maker into a supplier of AI compute, and the data confirms it. In the June quarter, data-center revenue hit $6.7 billion, more than double the year-ago figure, and it now accounts for 58% of total revenue. A year earlier it was just over two-fifths of the business. The company that used to make most of its money from chips inside laptops now makes most of its money from chips inside the largest AI data centers on the planet.

The driver is contracted demand. AMD has signed multi-gigawatt, multi-year agreements to supply AI accelerators to the biggest buyers — OpenAI for up to 6 gigawatts, Meta for up to 6 gigawatts, Anthropic for up to 2 gigawatts. These are not hopes; they are orders that convert vague AI optimism into a schedule of future revenue. Citi described the mechanism in plain terms: AMD is emerging as a "second source" to Nvidia in the GPU market. If AI compute is the scarce resource and Nvidia is the single bottleneck supplier, the buyers' need for a second source is a durable reason AMD's revenue stream exists. The upgrades are the analyst world catching up to exactly that fact.

The divide that decides

Put the two together and the picture resolves. The sellers after earnings were pricing the expectation; the upgrade-writers are pricing the re-rating. Both can be right, and the disagreement is where the useful question lives. The re-rating has already been bought: the 160% gain and the 74-times multiple are the market's payment for AMD becoming Nvidia's second source. What the price targets cannot tell you is whether AMD will actually execute the booked demand — whether the Instinct MI450 and its Helios rack-scale system, shipping in the second half of 2026, ramp on schedule and at margin, and whether the "up to" in those gigawatt deals becomes "will." That is the difference between a re-rating and a bubble. A re-rating happens once; execution has to keep earning it for the stock to compound.

A wave of upgrades is a weak reason to own a stock — it is momentum wearing a report's clothes. The real reasons to own AMD — the doubling data-center business, the booked multi-gigawatt orders, the rarity of a true second source to Nvidia — are real. But they are already in the price at 74 times earnings. The upgrades will keep flowing while the story holds, because the story is true. That does not make the next price target the question that matters. The number that matters is whether the gigawatts AMD has contracted become revenue on schedule and at margin. That is the only question a stock up 160% with 74-times earnings still needs to answer.