Lip-Bu Tan spent part of Intel’s earnings call on Oct. 29 explaining why customers were buying fewer personal computers. The explanation had little to do with Intel.
Customers, the chief executive said, had begun slowing purchases because of memory prices and the availability of memory supply. Shares fell about 28% the next day, the worst single session for the company since the 1970s. The reaction was to a set of numbers that described a cost problem arriving in the shape of a demand problem.
Fourth-quarter guidance called for revenue of $15 billion to $16 billion. Wall Street had been modeling roughly $17 billion. Gross margin was guided to 37%, down from 43%, and the company said it expects free cash flow to be negative for the period. Cuts of that size usually signal lost share or a collapsing market. This one was mostly about the price of a component Intel does not make.
That distinction shapes how the quarter should be read. Intel’s client business sells processors into machines whose bill of materials now includes memory at prices that have climbed for four consecutive quarters. When memory gets expensive, PC makers raise prices, cut configurations or ship fewer units, and each of those choices reduces the number of Intel chips ordered. Management’s own data-center revenue rose 78% from a year earlier, which is what the AI buildout looks like from inside a chipmaker that is finally shipping into it.
The margin line is harder to explain away. Six points of gross margin is not a demand effect. It is the arithmetic of running fabs below the utilization they were designed for, of bringing up new process nodes before they yield, and of holding prices competitive against a rival with a lower cost structure. Intel has been carrying all three at once for two years, and the fourth quarter is when the bill for them lands on the same page as a shrinking market.
Its answer to that has been capacity, and capacity keeps a calendar. The company told investors that output from its new fabs will not be available in volume until 2028. That is the year the server processor market is supposed to have grown enough for the capacity to matter. It is also two years of depreciation away from earning anything.
Analysts moved quickly. Wells Fargo cut its price target to $80 from $120 that morning, part of a broader reduction across semiconductor names after a difficult week of earnings. The stock had been valued on the premise that the foundry business would eventually sell capacity to outside customers at margins Intel has not yet demonstrated. A guidance cut driven by memory prices does not change that premise, but it does push the date on which the premise can be tested.
The political dimension is new. The U.S. government owns roughly 10% of Intel, a stake taken in 2025 to keep leading-edge logic manufacturing on American soil. On Oct. 30 that holding lost about $12 billion of value on paper. Industrial policy is designed to absorb that kind of mark-to-market without changing course, and there is no indication it will. Still, the state is now a shareholder in a company whose share price is being set by a memory market it cannot influence.
The strategic case has not moved. Intel remains the only American company attempting leading-edge logic manufacturing at scale, its customers include the federal government, and the demand forecast management offered — a server processor market growing several-fold within five years as inference migrates off accelerators — would justify the spending if it arrives anywhere close to schedule. The dispute is about timing, and about whether the company can fund the wait from operations rather than from the balance sheet.
There is an irony in the sequencing. Intel spent the past three years arguing that AI would eventually run on general-purpose processors rather than on accelerators alone, the chips that made Nvidia the most valuable company in the world. The first AI-driven shock to hit its own business did not come from that shift. It came from the memory chips that surround the accelerators, whose prices are now high enough to change what consumers pay for a laptop.
The company’s products are selling. The market for them is being constrained by something outside its control, at the same moment its manufacturing transition is consuming cash. Until memory prices ease or PC makers stop passing the increase along, Intel’s guidance will be set by a supply chain it does not own, and its shares will trade on the difference between what it can make and what its customers can afford to buy.
Management framed 2028 as the year the new capacity arrives. Investors heard a year that is two more annual reports away, after a year in which capital spending runs near $20 billion and customers are buying fewer of its chips. Whether that patience holds depends on yields at the new node and on whether outside customers commit to volume. Those numbers arrive quarterly. The next set is due in January, and the memory market will have a vote.
