3 AI-Dependent S&P 500 Stocks Now Flashing Warning Signs

Follow the money behind this year's stock market and it all points one direction: artificial intelligence. That single trade powered a historic run in semiconductors — and it has already cracked twice. The Philadelphia Semiconductor Index, the group's bellwether, fell into a bear market in July, down roughly 20% from its late-June high even while it still sat more than 60% higher for the year. Since then the tech-heavy QQQ has crossed the 10% pullback threshold traders call a correction.

A warning sign is not a forecast that a stock will fall. It is a measurable mismatch between what a stock costs and what it can reasonably earn — and whether that earning power is durable. Three AI-dependent S&P 500 names each show a different version of that mismatch. One's profits look cheap but may be peaking. One's price is far ahead of the profits it actually books today. One trades as if flawless growth is guaranteed. They are worth examining together, because sector-relative comparison is where cheapness, growth, and risk actually get defined — no stock means anything in isolation.

Micron: the cheap-looking cyclical peak

Start with Micron, the most striking case. The stock, trading near $979, is up more than 240% this year and occasionally moves 3% in a single session — the signature of a crowded, crowded trade. That is also roughly 22% below its 52-week high. The run was breathtaking; the warning lives in the earnings math that follows it.

Micron trades at about 22 times trailing earnings, which looks reasonable to a casual reader. But it carries a forward multiple closer to 155. Convert those multiples back into dollars and the gap becomes the story: the market is paying for roughly $45 a share of trailing profit yet pricing in only about $6 to $7 of forward earnings. That is the market telling you it expects a sharp rollover in the memory cycle after one of the most explosive quarters in company history — revenue near $23.9 billion and earnings per share around $12.20.

This is the classic cyclical trap: a low trailing P/E that is only cheap because earnings sit at a cyclical peak. Memory is notoriously mean-reverting, and the industry is already racing to lock in supply — Nvidia alone disclosed roughly $279 billion of supply and capacity commitments, up from $119 billion a quarter earlier, much of it tied to the high-bandwidth memory that powers AI servers. That is a bet on continued demand, but it is also a wall of capacity being built, the historical setup for oversupply once growth normalizes. Micron's balance sheet is fine — about $20 billion of net cash — so this is not a solvency story. It is a durability story: you are paying peak prices for peak earnings, and the forward curve says the peak does not last.

AMD: valuation ahead of the profit

AMD is the mirror image — a stock whose price has outrun the profits it actually earns. It is up about 141% this year and trades around 131 times trailing earnings and nearly 88 times EV/EBITDA. That second multiple, which strips out debt, cash, and taxes so companies can be compared fairly, is the useful one here: AMD is being valued at close to three and a half times Nvidia's EV/EBITDA while earning a fraction of what Nvidia earns on its capital — about 8% return on invested capital versus Nvidia's mid-80s.

None of this dismisses the business. Revenue grew about 40% year over year, and analysts expect the AI ramp to keep lifting it. But AMD's operating margin is only about 12% and its return on equity roughly 10% — the economics of a company still spending heavily to take share, not one that has already converted growth into profit. Paying 131 times earnings for a business earning single-digit returns on capital leaves essentially no room for the plan to wobble, and it pays no dividend to cushion the wait. The valuation assumes AMD converts its AI opportunity into profit at scale — an assumption its current income statement does not yet support.

Nvidia: perfection already in the price

Nvidia is the most recognizable name and the cleanest reminder that a warning sign can sit on top of an excellent business. Its fundamentals are superb: revenue up more than 80% year over year, an operating margin near 64%, and a balance sheet that prints cash. The question is what the market charges for that.

At a valuation near $5.3 trillion, Nvidia trades at roughly 17 times sales and about 58 times forward earnings — multiples that assume the AI buildout keeps compounding with few stumbles. That is not unreasonable for a company growing this fast. The warning is different: Nvidia has underperformed its own sector this year, up only about 17% while AI peers logged triple-digit gains, which means the market has already begun to de-rate it. Its near-term momentum has turned mildly negative over the past month. And the fragility shows at the index level: the July slide into the chip group's bear market was sparked by doubts that the heavy spending on AI will earn a return, after a Chinese model that runs on far fewer chips raised questions about whether America's buildout is as lucrative as the price tags assume. For a stock this size, held by this many funds, a wobble in that narrative does not just dent Nvidia — it shakes the whole trade.

What three different warnings add up to

Put them together and the picture is not that the AI trade is broken. It is that the easy, obvious part is priced. Micron asks you to accept that today's peak profits are tomorrow's base. AMD asks you to fund profits that have not appeared yet. Nvidia asks you to pay for flawless execution from the largest, most-crowded chip stock in the world. Those are three different risks, and the mistake would be treating them as one.

Warning signs are not sell orders — a rating cooling is the process working, not the business breaking. What they justify is structure. If you hold these names, know which of the three risks you are actually carrying, and hold them inside a plan that pairs concentrated growth with durable cash-flow businesses — the barbell — so a reset in the AI trade does not drag your whole portfolio with it. If you are watching from the sidelines, the useful question is not whether these are good companies. They demonstrably are. The question is whether the price you would pay today still leaves you a margin of safety. On the forward numbers, several already have it priced to one side.