Markets are closed today for Labor Day, which makes it a good moment to look at what the last week actually established — because the pattern was consistent enough to be worth naming.
Five large companies reported results between Wednesday and Thursday. Four beat expectations on both revenue and earnings. Three of the four fell. The one that rallied gave most of it back within twenty-four hours.
The scoreboard
Snowflake (SNOW) was the cleanest beat of the group. Fiscal Q2 product revenue grew 37% to $1.49 billion — the third consecutive quarter of accelerating product revenue growth — with adjusted EPS of $0.62 against a $0.45 consensus and net revenue retention of 126%. Management raised full-year product revenue guidance to $6.07 billion, up 36% year over year, from $5.84 billion previously. The company added 692 net new customers, up 32%, and CEO Sridhar Ramaswamy said Snowflake "continues to power the enterprise AI revolution."
The stock rose 16.55% Thursday to $356.47, a fresh multi-month high. Then on Friday it fell 5.41% to $337.18, with the S&P 500 down only 0.38% and enterprise-software peers roughly flat.
Nothing changed overnight. Guidance held, disclosures were routine, no analyst event of consequence occurred. What happened was that a one-session 17% gain met the natural supply of holders who had been waiting for exactly that gain to arrive.
Broadcom (AVGO) grew total revenue 86% to $29.6 billion and AI semiconductor revenue 221% year over year to $16.7 billion, with free cash flow of $13.7 billion — 46% of revenue. CEO Hock Tan guided Q4 AI semiconductor revenue to $21.7 billion, up 236%. The stock fell 2.74%, because Q4 total revenue guidance of roughly $34.8 billion sat marginally below a $35.03 billion consensus. The bar was already set well above what the company guided to.
Ciena (CIEN) posted the largest year-over-year improvement of the group: revenue up 37% to $1.67 billion, adjusted EPS of $2.11 against $0.67 a year earlier, adjusted operating margin of 22.5% versus 10.7%, and a raised full-year outlook of $6.42 billion. The stock fell 10.36%.
Zscaler (ZS) beat on both lines — revenue up 25% to $898.2 million, adjusted EPS of $1.19 versus $1.09 expected, ARR of $3.771 billion — but guided fiscal 2027 revenue growth to 16.6%–17.5%. That confirmed the step down from 25% growth rather than reversing it. Shares closed Friday at $169.80, down 4.50%.
lululemon (LULU) was the one genuine disappointment. Revenue fell 4%, comparable sales fell 9%, Americas comps fell 12%, and the company cut full-year revenue guidance to $10.35–10.50 billion from $11.00–11.15 billion. Reported EPS of $2.92 included $134.5 million of tariff refunds worth $0.86 per share. The stock fell 17.38% Friday to $100.61.
And the week before it, MongoDB grew revenue 30% — its fastest pace in years — raised full-year guidance, and lost roughly a seventh of its value the next morning.
What the pattern actually means
The obvious reading is that the market has turned hostile to technology. That reading is wrong, and the tape says so: the S&P 500 finished the week essentially flat, up about 0.1%, and closed Friday at 7,718.60 against 7,711.76 a week earlier. Broad indexes do not go sideways during a sector unwind.
The more accurate reading is narrower. Expectations have caught up with the fundamentals. For roughly two years, AI-exposed companies could clear the bar simply by growing fast, because the market had not yet priced in the growth. That gap has closed. A company growing 37% into a stock that already assumes 37% growth produces no return. A company growing 221% into a stock that assumes 236% produces a decline. The same dynamic showed up when AI revenue tripled and the stock still fell 9.4%.
This is not a warning about the underlying businesses. Ciena's optical revenue, Broadcom's custom accelerators and Snowflake's consumption growth are all real, all accelerating, and all funded by capital-expenditure commitments that extend years out. The spending is not slowing.
What has changed is that the reaction function has inverted. Beating a number is no longer sufficient. Beating the number embedded in the price is the only thing that pays.
For anyone holding these names, the practical consequence is that position sizing and entry price now matter more than being right about the industry. Being correct about AI infrastructure demand has been the consensus view for two years. Consensus views do not generate excess returns; they generate expensive entry points.
The macro turn underneath it
The week's other event reframed everything for the next two weeks.
August nonfarm payrolls, released Friday morning, came in at 162,000 against expectations near 55,000 — roughly triple the forecast, and above every estimate in Bloomberg's survey. The unemployment rate held at 4.1%, and June and July payrolls were revised up by a combined 55,000, turning July's reported 23,000 job loss into a 21,000 gain.
That single print moved CME FedWatch odds of a September rate hike from about 50% to roughly 60%, sent the two-year Treasury yield to 4.38% — its highest since January 2025 — and knocked gold down more than $125 before it recovered part of the move.
The counterargument is in the wage data. Average hourly earnings decelerated to 3.1% year over year, the slowest since May 2021, and the labor force expanded by 683,000 people. Strong hiring with cooling wages is expansion, not overheating.
But Fed Chair Kevin Warsh has been explicit that inflation, not employment, is his binding constraint, with twelve-month PCE at 3.7%. Friday's report removed the "the labor market is too weak to hike" argument. That leaves the inflation data to decide it.
What lands next
Markets reopen Tuesday. The shortened week carries the two prints that determine the September 15–16 FOMC meeting:
Thursday, September 10: August PPI, followed after the close by earnings from Oracle and Adobe
Friday, September 11: August CPI — the final inflation reading before the Fed decides
A hot CPI on top of a 162,000 payroll print and an ISM services prices index at a four-year high makes a hike the base case. A cool one reopens the debate entirely.
Four trading days, two data points, and a Federal Reserve meeting immediately on the other side of them.
This article is for informational and educational purposes only and does not constitute investment advice or a recommendation to buy or sell any security.
Found this helpful? Share it with others.