Analysis
Three AI-exposed companies reported earnings within 48 hours of each other this week, all three beat expectations, and the stock market treated them in opposite directions:
- Salesforce -- rose roughly 23% on Aug. 27 after Agentforce and Data 360 combined ARR hit nearly $3.9 billion, up more than 210% year over year, with Agentforce ARR alone crossing $1.5 billion on 240%-plus growth, The Motley Fool reported
- CrowdStrike -- jumped roughly 20% the same week after record net-new ARR of $333 million and 25% ARR growth to $5.84 billion, driven by what CEO George Kurtz called AI-security demand
- Marvell -- beat its own quarter, raised its fiscal 2028 revenue target to roughly $18 billion from $16.5 billion, implying 50% growth, and fell 6% anyway, a divergence Pulse covered in detail as it happened this week
All three beat. Only two got rewarded.
“## What actually separates the winners from the loser All three companies beat consensus.”
What actually separates the winners from the loser
All three companies beat consensus. The difference is what the beat is made of. Salesforce and CrowdStrike's growth is customer-level ARR -- named enterprise deployments generating recurring revenue today, verifiable against signed contracts and renewal cohorts. Marvell's growth is a forward guidance number for custom AI chips that mostly haven't shipped yet, dependent on hyperscaler capital-expenditure plans holding steady for two more years. The market has apparently decided it will pay a premium for AI revenue it can already see, and a discount for AI revenue it has to trust a multi-year forecast to deliver -- even when the forecast comes with a formal guidance raise from management.
The wider pattern
This isn't isolated to one week. Nvidia itself reported a record $96.2 billion quarter days earlier and still rose nearly 9%, which on its face contradicts a hardware-skepticism thesis -- except Nvidia sits at the top of the compute stack with pricing power no other chip supplier has, while Marvell is one layer down, selling custom silicon into hyperscaler capex budgets it doesn't control. Broadcom, the larger incumbent in Marvell's own custom-ASIC category, has shown the identical pattern before: sold off on results, then made new highs within weeks once the guidance held. Pulse covered Salesforce's own AI push building toward this quarter, including its Anthropic partnership work, and CrowdStrike's record quarter as it printed -- both moves that this week's price action now validates.
The counterweight
Reading three days of stock moves as a durable rule is exactly the kind of overreach that carries real risk of being wrong within a month. Marvell's post-earnings drop came after the stock had already rallied hard into the print -- a classic sell-the-news pattern that has nothing to do with the durability of its custom-silicon thesis. And software ARR is not immune to its own version of Marvell's problem: Agentforce and Data 360's 210% growth rate is measured against a small base, and CrowdStrike's $333 million in net-new ARR still has to retain and expand, not just land. A software beat that misses next quarter's expansion number will get the same treatment Marvell just got.
The number worth tracking through the next earnings cycle is gross margin trajectory at Marvell as its ASIC mix rises against ARR retention at Salesforce and CrowdStrike as their AI products mature past the initial adoption spike -- whichever holds up first tells you which side of this divergence was actually right.
