Two of the loudest names in the AI trade report after the bell tonight, and the setup is brutal. Marvell Technology is up 131 percent year to date heading into its Q1 print. Salesforce is being asked to prove that “agentic AI” is a real revenue line and not a buzzword the CFO repeats on calls. And both companies just watched Zscaler beat earnings on Tuesday and lose 17 percent of its market cap by the next open.
That is the bar tonight. Not “did you beat.” More like, “did you beat by enough.” Look, that is a brutal way to walk into a print. Both CEOs know it.
The Marvell setup
Marvell trades around $196 going in, after roughly doubling in 2026. Wall Street wants $0.80 EPS on $2.40 billion in Q1 revenue, about 27 percent year over year growth. Data center is 75 percent of the business now. Custom silicon for hyperscalers, what Marvell calls XPUs, is the story everyone is here for.
If you have been following the AI chip narrative, you know the pitch by now. Amazon, Microsoft, Google, Meta all want their own AI chips. None of them want to design a whole chip from scratch. Marvell does the design work, ships them the silicon, collects the revenue. Twenty plus custom AI chip wins are already locked for FY2028 and FY2029 production.
The Street walked into this print pre warmed. HSBC upgraded MRVL to Buy on Tuesday citing a stronger cycle in AI networking, sending the stock up over 6.5 percent in premarket. The bull case is that data center spend is still accelerating, optical interconnect demand is underestimated, and Marvell sits in the middle of all of it.
Here is the thing nobody on the bull side wants to say out loud. A handful of hyperscalers control the entire custom silicon book. Lose one, and the FY2028 revenue ramp gets ugly fast. Wall Street is comfortable assuming that does not happen. Tonight is when Marvell either confirms that confidence or feeds it doubt.
What Marvell investors actually want to hear
What Wall Street wants tonight is not just the beat. They want CEO Matt Murphy to talk about FY2027 like the custom AI ramp is happening on time, with no slip, no margin compression, no surprise customer concentration risk. Anything less and the stock prices in what it has already priced in: a perfect future.
Specific things to watch on the call. Is the FY2028 design win count still growing, or is it stuck at twenty plus. Are gross margins holding at or above 60 percent, because custom silicon is supposed to be a higher mix story. Is data center revenue still 75 percent of the pie, and is it growing sequentially, not just year over year. Any one of those metrics turning soft is the kind of thing that turns a 5 percent beat into a 10 percent after hours drop.
The Salesforce setup
Different stock, same problem. CRM is being asked to grow 12 to 13 percent on top of an enormous base. Wall Street wants $3.12 EPS on $11.05 billion, per Jefferies’ preview note circulated Tuesday. That is a $45 billion plus run rate company trying to convince the market that Agentforce, its AI agent platform, is the next pillar.
Agentforce ARR is up 169 percent year over year to about $800 million, with Agentforce and Data Cloud together adding $2.9 billion in ARR by the close of fiscal 2026. Impressive number, until you remember Salesforce’s total revenue is north of $40 billion. So Agentforce, right now, is roughly 2 percent of the business. The bull case is that 2 percent becomes 15 percent in three years. The bear case is that every SaaS vendor on earth, plus Microsoft Copilot, plus a thousand AI startups, are charging into the same use case at the same time.
CEO Marc Benioff has been calling Agentforce the “third wave” of Salesforce. The market has not decided yet whether that is vision or sales pitch. Tonight is when he has to show some receipts.
The Agentforce credibility problem
The competitive context for Salesforce is worse than the topline suggests. CRM stock is down 32 percent year to date. Bank of America has a $160 target. The bulls see meaningful upside above $260. That spread is not a forecast, it is an argument.
Honestly, the part that worries me is the seat math. A quarter of Salesforce’s partner survey respondents reported seat reductions at renewal, up from 5 percent in the prior quarter. The bear thesis is straightforward. AI agents replace seats. Seats are how Salesforce gets paid. If the agent revenue does not grow faster than seats erode, the model breaks. Tonight is when the market gets its first real data point on whether that math is starting to bite.
Why Zscaler is haunting this print
Here is the line that has every CFO on Sand Hill Road sweating today. Zscaler reported Tuesday night with revenue of $850 million, up 25 percent year over year, EPS of $1.08, and a raised full year guide. The company talked up its position as the cybersecurity platform for the AI era. The stock is down roughly 17 percent on Wednesday.
What went wrong? Revenue came in just shy of the $860 million Wall Street was expecting. A miss of barely over 1 percent on the top line. In a normal market, that gets shrugged off. In an AI bull market where multiples are 60 times forward earnings, that gets you crushed.
The lesson for tonight is uncomfortable. The bar for any AI exposed name is not “good.” The bar is “perfect on revenue, perfect on guide, no excuses on margin, no rambling on the call.” Marvell and Salesforce are both walking into that meat grinder after the bell.
The $145 billion question
Step back from the two prints for a second. Meta raised its 2026 capex range to $125 to $145 billion this spring, almost double last year. Microsoft, Google, Amazon all spending at record clips. Nvidia’s print last Wednesday blew past $78 billion in revenue and the AI capex story stayed intact.
Marvell is the most direct downstream beneficiary of that spend after Nvidia itself. If Marvell guides cautiously tonight, the market will not read it as a Marvell story. It will read it as a hyperscaler capex story. And that is the one piece of the AI thesis the bulls cannot afford to see crack.
Why This Matters
If both Marvell and Salesforce print clean and guide up, the AI rally that has carried the Nasdaq to 26,343 has another leg to run. Hyperscalers keep buying custom silicon. Enterprises keep paying for AI agents. The trade works.
If either one misses or guides cautiously, you get a repeat of the Zscaler tape and a lot of nervous fund managers heading into June. Marvell is the more dangerous print, because a custom AI chip miss would be read as a signal that hyperscaler capex is finally rolling over. That is the one thing the bulls do not want to hear after Meta already pushed its 2026 capex range up to $125 to $145 billion this spring.
This is not just two stocks. It is the entire AI capex thesis getting a real time check.
USABlaze Takeaway
The AI trade has been priced for perfection since February. Tonight tests whether anything less than perfection still gets a polite shrug from the market, or whether we are sliding into the part of the cycle where good is no longer good enough.
Watch the after hours tape on MRVL and CRM at 4:30 pm ET. If both are green and holding, the Memorial Day rally extends. If one is down five percent and bouncing around, the conversation tomorrow morning is about which AI name reports next and how scary that print suddenly looks.
My read: Marvell beats and holds, because the hyperscaler demand is structurally there and HSBC’s upgrade walked in at exactly the right time. Salesforce is the riskier print, because Agentforce has to grow into a number the market has already priced in. If CRM disappoints, the seat-erosion thesis becomes the entire software-sector conversation for June. The Zscaler chart on Wednesday is the warning sign nobody is supposed to be looking at. Tonight we find out if it was a one off, or the start of something.
Sources: Yahoo Finance (Marvell), Yahoo Finance (Salesforce), The Globe and Mail, The Motley Fool (Zscaler). Featured photo by Immo Wegmann via Unsplash.
By The USABlaze Editorial Desk