Dan Ives says the semiconductor sell-off is "way overdone." The Wedbush analyst delivered that verdict as fears over AI capital spending rattled four of the sector's most exposed tickers: Nvidia (NVDA), Micron (MU), SK Hynix (SKHY) and Sandisk (SNDK).

The statement matters because of who said it, not merely what was said. Ives has been among the most prominent Wall Street voices on AI infrastructure spending, and his willingness to call the decline an overreaction puts a number on sentiment, or at least a label on it.

Why are NVDA, MU, SKHY and SNDK falling together?

The four names share one exposure. Nvidia supplies the GPUs at the center of AI data center buildouts. Micron, SK Hynix and Sandisk supply the memory that sits next to those GPUs, chiefly high-bandwidth DRAM and NAND.

When investors begin to doubt the durability of AI capital spending, the whole chain sells off at once. A cut in hyperscaler budgets would hit GPU shipments first and memory demand second. The market, in its usual efficient panic, prices both simultaneously.

That coupling explains the breadth of the decline. This is not a company-specific story about any one of the four. It is a story about whether the AI infrastructure cycle has a ceiling, and whether the market has finally decided to test that hypothesis.

What does Ives actually dispute?

Ives's argument, condensed to its core, is that the sell-off misprices the underlying demand. His phrase "way overdone" implies the market has moved prices further than any plausible deterioration in fundamentals would justify.

The bear case rests on AI spending fears. The worry is straightforward: hyperscalers have spent enormous sums on AI data centers, and at some point returns must appear. If they do not, budgets contract and the component suppliers feel it first.

Ives positions himself against that timing. The disagreement is not about whether AI spending will eventually normalize. Every capital cycle ends. The disagreement is about whether the current correction reflects confirmed weakness or mere anxiety.

So far, the correction has been driven by sentiment rather than by reported order cancellations or revised shipment guidance from the companies involved. That distinction is the crux of the analyst's case. A sell-off built on fear leaves more room for recovery than one built on filings.

Who bears the risk if the bears are right?

The memory makers carry particular exposure. Micron, SK Hynix and Sandisk have leaned heavily into AI-driven demand, and their pricing power has rested on tight supply of AI-grade memory. A demand shock would compress margins quickly in a business that has never needed much help being cyclical.

Nvidia's risk is different in shape but similar in direction. Its valuation presumes sustained acceleration in AI infrastructure spending. Any credible signal of a slowdown, from any large customer, would pressure the stock harder than most.

The four-way sell-off also illustrates how concentrated the AI trade has become. One thesis, expressed across four tickers in two countries and two market segments. Diversification within semiconductors has become something of a courtesy.

What happens next?

The resolution will come from evidence, not commentary. Upcoming earnings reports from the affected companies, plus capital expenditure guidance from the major hyperscalers, will either validate the spending fears or confirm Ives's read that the decline outran the facts.

Until then, the analyst verdict stands as a marker. The sell-off is "way overdone" in his judgment, and the market now has a number of weeks to decide whether he is early or simply wrong. In semiconductors, that question has historically been settled without much gentleness.