Ben had started checking his retirement spreadsheet before checking his work email. He was a software engineer, and on good market days the sheet suggested he might stop being one earlier than expected. He pictured a slow weekday breakfast with no meeting waiting behind it. Recently he had moved his possible retirement date forward without lowering the growth rate he was counting on.
Some of his savings were in companies linked to AI, including businesses selling the hardware behind it. He had paid particular attention to memory-chip and storage companies. An August 6 Reuters report said Sandisk shares had risen more than fivefold since the start of the year. Ben had bought later, so he had missed much of that rise. Still, he had raised the growth rate in his retirement sheet after watching these stocks.
He had tried to prepare himself. Reminiscences of a Stock Operator sat beside Poor Charlie's Almanack on his desk, with slips of paper marking passages he wanted to remember. He read them on quiet weekends, away from the prices on his work screen.
On Wednesday, October 28, 2026, Alphabet released its results after the regular US trading session. Its shares fell in extended trading. By Thursday's close, the Class A shares had lost about 12.6% for the day. Amazon and Intel had fallen during Thursday's regular session too. Ben watched the red prices and wondered whether investors were giving up on AI-related shares.
After Thursday's close, Amazon and Intel published their own results, and both fell further in after-hours trading. Ben had been waiting for dinner to cook. He turned the heat down and opened Yahoo Finance instead, searching for why the shares were falling. He wanted someone to tell him whether this was the top. One story led with weak sales. Another focused on the cost of new equipment.
He had read Amazon's falling price as a reaction to its results. Then he noticed when the report had appeared. Alphabet's report came on Wednesday evening. Amazon's and Intel's came on Thursday evening, after their regular-session declines had already happened. Those earlier falls could not have been reactions to reports that had not yet been released. He put a line between the regular close and after-hours prices in his notes.
The first Alphabet figure made him check that he had opened the right report. Quarterly revenue had risen 25% to about $127.6 billion. Google Cloud revenue had more than doubled from a year earlier to about $30.7 billion. These were sales for the quarter ending September 30. He went back to the headline, unsure what he had missed.
Yahoo Finance's earnings coverage gave him the missing comparison. The analyst forecast it reported for Alphabet's total revenue was about $127.9 billion. The result fell short. Advertising sales were also below the forecast, although Cloud beat expectations. Ben looked again at the yearly growth figure. It did not show what investors had expected.
Then he reached the cash figures. Alphabet generated about $42.3 billion of cash from operating activities during the quarter and paid about $51.4 billion for property and equipment. Subtracting the second amount from the first gave negative free cash flow of about $9.1 billion. Alphabet still reported an operating profit, but its equipment payments had used more cash than operations brought in.
He had recently read about what goes inside an AI data center and what the hardware costs. This time the spending felt less distant. The equipment might earn money over several years, but paying for it used cash now. Ben could see why shareholders would ask when the investment would pay back. He could not tell how much of the equipment spending was for AI. He would also have to account for other Cloud services before treating its growth as a measure of AI sales.
Ben opened Amazon's report, still thinking about how quickly those equipment purchases would produce returns. Quarterly sales at its AWS cloud business were about $46.0 billion, up 39% from a year earlier. Reuters said AWS beat analyst expectations, while Amazon's total sales fell slightly short. AP reported that earnings per share had missed expectations too. The cloud business was growing at both companies, and the share prices were falling anyway.
Amazon reported negative free cash flow of about $42.4 billion for the twelve months ending September 30. Higher investment in property and equipment drove the change. Ben read the dates beside the figure. This covered a whole year, whereas Alphabet's negative $9.1 billion covered one quarter. He could not compare the amounts directly.
He scrolled to Amazon's forecast for the holiday quarter. The company expected sales between $240 billion and $247 billion. Reuters put the LSEG estimate at about $244.5 billion, inside that range, though above its midpoint. That left a possibility he had missed in the headline. Sales near the top of Amazon's range would beat the analyst estimate. Ben went to stir dinner while the earnings page was loading. He had left it too long.
Intel's story was harder to fit into his first theory. Its adjusted earnings for the completed quarter beat the analyst estimate Reuters cited. Its forecast for the next quarter was weaker than analysts expected. Intel expected revenue of $15 billion to $16 billion against a LSEG estimate of about $17.1 billion. Intel's whole forecast range was below that estimate.
Reuters also described weak demand for personal computers and disruption from an explosion at an Intel factory in Israel. Higher memory costs were making PCs more expensive, which could hurt demand for machines containing Intel processors. Ben paused at that detail. A memory supplier could benefit from higher prices while a PC business struggled with the resulting costs.
There was a similar tension in the data center spending news. Google was paying for equipment that suppliers wanted to sell. Its spending created orders for other companies, so a cut could help Google's cash flow but hurt suppliers' sales. Yahoo's reporting on Thursday described concern that big technology companies' capital spending had come below market estimates. Spending could rise from last year and still disappoint a market expecting even more.
He returned to the storage stocks he had watched rise earlier in the year. A Micron annual report described how changes in supply and demand could move memory selling prices and profits. Strong demand could lift prices while manufacturers struggled to add supply. More production later could change that balance, especially if customers slowed their buying. A factory planned during strong demand might open after customers had reduced their orders.
Ben had treated a useful product as a reason for an endlessly rising stock price. Yet a memory company could sell more storage and still face falling prices per unit. Investors paying for large future profits could lose money if those profits disappointed them. Sandisk sold NAND flash storage, which was different from the high-bandwidth memory used beside some AI processors. He had run into that difference when reading about the storage and memory needed to run AI locally. Now he checked the product descriptions before grouping the stocks together.
By then he had found real reasons for concern. The companies were committing large amounts of money, and shareholders could reasonably question the returns. Intel's forecast contained a business warning he could not wave away. But the strong Cloud and AWS figures did not support his first picture of AI demand suddenly disappearing everywhere. None of the reports told him whether Thursday marked the final peak of a stock cycle. He refreshed the prices again without finding anything that settled it.
He reached for Poor Charlie's Almanack, then left it closed. Remembering a sensible passage would not tell him how much of his retirement depended on the same kind of spending. He opened his holdings instead. Different company names had made the account feel spread out, but several positions relied on continued demand for AI infrastructure. A slowdown in that spending could affect more than one of them.
On FINRA's site, he found an explanation of concentration risk and a reminder to look inside funds too. He opened the holdings list for a fund he owned and searched for names already in his account. Before deciding what to change, he wanted to add up the overlapping positions. He had not done that when he bought the fund.
He tried a 20% fall in his retirement sheet. Every $100,000 would become $80,000, and recovering that loss would take a 25% gain. He had not built that fall into his earlier version. If he also needed to sell investments to cover living costs during the fall, less money would remain invested for a possible recovery. He wrote down his own expected expenses without assigning them a guaranteed future stock return.
The retirement sheet stayed open beside his work email. Ben removed the assumption that recent gains would repeat and saved separate versions with weaker returns and different retirement dates. Beside his expenses he added a note for an investment adviser. He wanted to ask how he could cover those costs if his shares fell just as he retired.
He finally scraped dinner out of the pan and took the plate to the table. The laptop stayed on his desk with the retirement sheet open. He could read the dates from where he was sitting.