The most useful fact about this week is that Datadog beat revenue estimates by 4.3%, beat on earnings by nearly 38%, raised full-year guidance, guided the third quarter 3.2% above consensus, and lost 15% in a single session. That is close to the best quarter a software company can deliver, and the market took it apart anyway. Nothing about this week is explained by weak results. Results were excellent almost everywhere. What changed is the price of an excellent result.
Start with the counts. Roughly 440 S&P 500 companies have now reported, with 87% beating against an 82% beat rate a year ago, and average year-over-year earnings growth running near 25%. Beats have stopped being information. When almost nine in ten companies clear the bar, clearing the bar tells an investor nothing about relative positioning, and the market has adjusted by moving the question from whether a company beat to whether the beat contained acceleration the buyer had not already paid for.
AMD is the cleanest illustration. Second-quarter revenue came in at $11.54 billion, up 50%, with data center revenue at $6.7 billion, up 107%. Non-GAAP earnings of $1.66 rose 246% from a year ago. Management guided the third quarter to roughly $13 billion at the midpoint, about 41% growth, and confirmed that Helios systems begin shipping this quarter. Every headline number cleared. The stock fell 8% to 9% after hours on a single line: GAAP gross margin of 54% against a 56% consensus, with the shortfall attributed to near-term Helios ramp costs. AMD had more than doubled in 2026 and rallied 7.8% into the print. Options had priced a 7.8% move in either direction, and the move arrived on the downside for a two-point margin variance in a quarter that grew data center revenue by more than 100%.
That is the shape of the week. Revenue is no longer contested in the AI complex. Demand is not the argument. The argument has moved entirely to margin and to the cost of servicing that demand, which is a much harder thing for management teams to guide and a much easier thing for the market to punish.
SanDisk delivered the week’s most extreme version of the same problem. Fiscal fourth-quarter revenue of $8.97 billion came with adjusted earnings of $39.25 against a $33.38 estimate, a 17.6% beat, on a gross margin of 84.6% and 51% sequential revenue growth. Shares closed down 5.4% at $1,350.50 and fell another 4.2% after hours. The reason sits inside the company’s own disclosure: approximately $2.01 billion of the sequential revenue growth came from pricing. That is a quality-of-earnings question the market had been content to ignore for three quarters and suddenly decided to ask. Pricing-driven revenue capitalizes at a lower multiple than volume-driven revenue because pricing mean-reverts and capacity eventually arrives. SanDisk went into the print with 25 buy ratings, five holds, and no sells. It came out of the print with a forward revenue midpoint below consensus and a shareholder base being asked to distinguish between a supply shortage and a business. Western Digital carried an implied move of nearly 12% into its own report after rallying 181% year to date and trading 62% above its 200-day moving average, which is a positioning problem regardless of what the numbers said.
Palantir shows what the market will still pay for. Revenue of $1.94 billion beat a $1.80 billion estimate and grew 93%, with U.S. commercial revenue up 149% to $764 million and U.S. government revenue up 90% to $809 million. Net income reached $1.07 billion against roughly $329 million a year earlier. Management raised full-year U.S. commercial guidance to more than $3.42 billion from $3.22 billion, and remaining U.S. commercial deal value more than doubled to $6.24 billion. The stock surged. The relevant context is that Palantir had lost 29% in 2026 going into the release. It was not priced for perfection. It was priced for deceleration, and it delivered acceleration. That asymmetry, not the growth rate, is what produced the move.
The same asymmetry explains Airbnb rising 17.4% on Friday to lead the S&P after a revenue beat that would have been unremarkable in a different setup, Magnite gaining 18% on a clean beat, and Shopify’s quarter being received as a monster. It also explains AppLovin falling 19.3% on its first guidance miss since going public, and Honeywell Aerospace dropping 21.4% after cutting 2026 organic sales growth guidance to 4% to 5% from 7% to 9%, a cut management attributed to supplier constraints rather than demand. In every case, the direction of the stock tracked the gap between the print and the positioning, not the gap between the print and the business.
Outside technology the tape was calmer and the results were mostly fine. Eli Lilly rose on strong GLP-1 demand and a raised full-year revenue outlook. Disney’s parks business strengthened. Uber came in roughly in line and slipped on third-quarter guidance below expectations. McDonald’s was soft. Warner Bros. Discovery beat on earnings and missed on revenue with streaming adjusted EBITDA close to doubling. SpaceX grew revenue 92% to $7.8 billion in its first public quarter and still fell toward $110 from above $135, though that has as much to do with the August 6 unlock of roughly 911 million shares as with the quarter.
The index level hid all of it. The S&P rose 3.6% on the week, the Nasdaq 5.2%, and the iShares Semiconductor ETF more than 7%, with both major indexes closing at records on Friday. That performance came from cooling oil on Iran talks and from a soft July payrolls print that removed the September rate hike from the front of the curve. It did not come from earnings. Underneath the record close, the highest-multiple names in the market spent the week absorbing double-digit single-day drawdowns on good news.
Which brings the week back to its defining number. Datadog raised guidance and guided above consensus and lost 15%. If a 3.2% guidance premium is now insufficient, then the operative question for every remaining Q3 report is not what the company will earn. It is how much acceleration is already embedded in the price, and there is no line in a press release that discloses it.
