Palantir’s shares jumped 29% on Tuesday, a day after the software company reported quarterly revenue up 93% and raised its full-year forecast by $500 million. Hours earlier, AMD had reported record revenue and profit that topped Wall Street’s estimate, and its stock fell as much as 9% after hours. SpaceX, reporting its first results since its public listing, saw its shares drop more than 8% after hours despite revenue that grew 92% and came in well above expectations. A week before, Intel had delivered its fastest revenue growth since the third quarter of 2011 and watched its shares slide about 7.5% the next day.
Four companies, four beats, four very different reactions. The pattern, investors said, says less about the companies than about the market that prices them.
The message of this earnings season, delivered in 48 hours by the four AI-linked companies that reported between Aug. 3 and Aug. 4, is that beating the number is no longer enough. What matters now is the distance between the result and the expectation that was already priced into the stock.
Palantir showed the logic working in both directions. Revenue of $1.94 billion, up 93%, was the fastest growth in the company’s history. U.S. commercial revenue jumped 149% to $764 million, approaching the company’s government business in size. Net income topped $1 billion for the first time. The company raised its full-year guidance by $500 million and its third-quarter forecast came in more than $160 million above consensus. Investors who had positioned for a blowout got one, and the stock rewarded them.
AMD offered the mirror image. Revenue of $11.54 billion was up 50%, a record, and came in above the $11.3 billion that analysts expected. Non-GAAP earnings per share of $1.66 beat the $1.62 consensus, and gross margin reached 56%. Data-center revenue more than doubled. Yet the stock fell 5.5% to 9% after hours. The reason, analysts said, was the outlook: the company’s guidance for the coming quarters, while in line with consensus, left no room for the kind of “super beat” that the share price had come to imply. Investors had pushed the stock up ahead of the report on hopes that the data-center boom would translate into another beat-and-raise. The actual number delivered exactly that, and it was still not enough.
SpaceX told a similar story on a grander scale. Revenue of $7.81 billion grew 92% and beat the $6.9 billion that analysts had modeled. The net loss narrowed to $541 million, and adjusted EBITDA of $3.54 billion was nearly double the $2 billion expected. But the company’s AI business, while growing revenue 247%, still burned $1.26 billion in operating cash terms, and investors who had bought the stock on the promise that AI infrastructure would turn profitable quickly sold it, sending shares down 8% after hours and as much as 12% at the next day’s open.
Intel completed the pattern. Its revenue growth was the fastest since the third quarter of 2011, a headline any turnaround CEO would envy. The stock fell anyway, because the growth came off a depressed base and the market had already priced a rebound in advance.
The common thread, analysts said, is that in a market where the biggest AI names trade at multiples that assume flawless execution, the bar has moved from “beat” to “beat by a wide margin.” Stocks now react not to whether a company cleared the consensus number but to how far the surprise overshot the informal targets that circulate among investors in the days before a report. Traders call them whisper numbers, and they have become the real benchmark.
The mechanics are straightforward. Sell-side analysts anchor their estimates to company guidance, which companies set conservatively. Options markets, meanwhile, price in moves of 8% to 12% around earnings for high-multiple AI names. When a stock’s options imply a double-digit move and the company delivers a single-digit beat, the report is, in effect, a miss relative to what the market already knows. The result is that good news and bad news travel at different speeds: a clean beat-and-raise can still fall short of a whisper number that had crept higher.
Analysts said the dynamic reflects a market that has matured. Two years into the AI spending boom, the easy comparisons are gone, and growth rates of 90% become harder to repeat. Investors are demanding evidence that demand translates into profits, and the companies themselves are split between the two poles: Palantir and AMD, which sell AI, are making money, while SpaceX, which builds AI infrastructure, is spending it. Karp’s “sovereign AI” framing and SpaceX’s AI losses capture the divide.
The next test comes in the coming weeks as more companies report and Nvidia, the sector’s most important name, gives its outlook. If the pattern holds, expect more volatility around reports regardless of the numbers inside them. For investors, the lesson of this earnings season is that the figure on the page matters less than the figure in the market’s head.
