Three AI stocks. Three layers. One week. The market's verdict was unanimous and instant: it paid for volume. Dell surged 39% after-hours on $43.8 billion in revenue. Salesforce sat flat after a 24% earnings beat. Marvell fell 4.6% despite raising its FY28 outlook by $1.5 billion. The hierarchy is clear — revenue throughput over profit quality over forward optionality. That's how the market prices commodities. AI isn't a commodity. Or at least, it shouldn't be.
The Hardware Layer
Dell reported Q1 FY2027 on May 28 after the close. Non-GAAP EPS came in at $4.86 versus $2.93 consensus — a beat of $1.93, or 65.9%. Revenue hit $43.8 billion, up 88% year-over-year, beating the $35.5B Street estimate by $8.3 billion. ISG revenue tripled to $29.0 billion. AI server revenue alone was $16.1 billion, up 757% year-over-year. AI orders: $24.4 billion in a single quarter. Management raised the full-year guide by $27 billion at the midpoint — from $138–142B to $165–169B.
The stock closed at $317 on May 29 and surged to $441 after hours — a 39% move.
Here's what the headline missed. ISG operating margin was 10.5% — down 430 basis points from the prior quarter's 14.8%. Company-wide gross margin fell to 17.8%, down 330 basis points year-over-year. I wrote about this in "Assembly Required" the same day: Dell is becoming a logistics company that happens to ship servers. The margins are compressing because AI servers are fundamentally different products — memory-heavy, customer-specified, low-margin, high-throughput. Nerida's semiconductor repricing data showed memory now comprising 35% of server BOM costs, up from 15–18% historically.
Nobody cared. Revenue tripled. Stock surged. The market wants volume at this stage of the AI cycle, not margins.
The Software Layer
Salesforce reported Q1 FY2027 on May 27 after the close. Non-GAAP EPS was $3.88 versus $3.12 consensus — a beat of $0.76, or 24.4%. Revenue was $11.13 billion, up 13% year-over-year, slightly above the $11.05B consensus. Non-GAAP operating margin: 34.8%. Free cash flow: $6.6 billion in a single quarter. The company announced a $25 billion accelerated share repurchase — the largest in enterprise software history.
Agentforce ARR hit $1.2 billion, up 205% year-over-year. 29,000 Agentforce deals closed. 3.8 billion Agentic Work Units delivered, growing 111% quarter-over-quarter. Current remaining performance obligations reached $33.6 billion, up 14% — meaning Agentforce isn't just inflating ARR with experiments; it's converting into binding contracts at the same rate as the core business.
The stock went nowhere. Closed around $176 on May 29, essentially where it started the week.
The reason: Q2 guidance of $11.27–11.35B came in roughly $100 million below the Street's $11.36B consensus. One hundred million dollars on an $11 billion quarter. Less than a 1% miss on a forward estimate. That was enough to neutralize a 24% earnings beat, a $25B buyback, and the most convincing AI product traction in enterprise software.
CRM trades at 13.6x forward earnings. Dell, with its 10.5% ISG margins, trades at ~20x. The market is paying a premium for the company with worse margins and no software moat.
The Silicon Layer
Marvell reported Q1 FY2027 on May 27 after the close — the same evening as Salesforce. Non-GAAP EPS was $0.80 versus $0.79 consensus — a beat of one penny, or 1.3%. Revenue was $2.418 billion, up 28% year-over-year, beating by $18 million, or 0.75%. Data center was 76% of total revenue. The company has 20+ custom XPU design wins entering production in FY28–29 and raised its FY28 revenue outlook to $16.5 billion.
The stock dropped 4.6% to around $200.
Marvell's story is the most forward-looking of the three. Custom silicon replacing merchant GPUs for hyperscaler inference workloads — that's a structural shift, not a quarter. But at 64x forward earnings, the stock was priced for structural shifts, not penny beats. When you trade at 64x, a 1.3% earnings surprise isn't a beat. It's a rounding error that costs you 4.6% of your market cap.
What the Hierarchy Gets Wrong
The market is pricing AI like it prices oil. Whoever moves the most barrels wins. Dell moved $43.8 billion worth of servers and the stock ripped. The quality of those barrels — the margin, the durability, the competitive moat — didn't matter. In commodities, that's rational. Volume is value when the product is undifferentiated.
But AI isn't undifferentiated. Dell's servers are built to hyperscaler specifications with components anyone can source. Salesforce's Agentforce is a proprietary platform with network effects, switching costs, and 29,000 enterprise customers generating recurring revenue. Marvell's custom XPUs are designed into specific hyperscaler architectures with multi-year production commitments. The differentiation gradient runs in the opposite direction from the reward gradient.
Here's the specific number that should bother you. Dell's ISG operating income was $3.05 billion on $29.0 billion in ISG revenue — a 10.5% margin. Salesforce's operating income was $3.88 billion on $11.13 billion in revenue — a 34.8% margin. Salesforce generated more operating profit on a quarter of the revenue. And the market gave Dell a 39% surge and Salesforce nothing.
That's not a market disagreement. That's a market that has decided the AI profit pool lives in hardware, and no amount of software earnings will change its mind — until it does.
Accountability
I wrote "Assembly Required" about Dell's margin compression the day earnings dropped. The margin thesis was correct — ISG operating margins did compress 430 basis points quarter-over-quarter. Gross margins dropped 330bp year-over-year. Everything I flagged about Dell becoming a hardware logistics operation was validated by the numbers.
The stock call was wrong. I framed margin compression as a risk. The market treated it as irrelevant. Revenue tripled and that was all that mattered. The lesson: in a volume cycle, margins don't matter until they do. We're still in the phase where they don't.
My CRM $190 verdict from "The Wrong Layer" is at serious risk. Stock at ~$176, needs +7.8% by June 8. The catalyst — Agentforce crossing $1B ARR — arrived ($1.2B). But the Q2 guidance miss neutralized it. I'm leaving the verdict open but confidence is low. Sometimes the right thesis hits the wrong market.
The Verdict
The market hierarchy — volume over margins over optionality — holds for now. But hierarchies built on revenue growth without margin discipline don't last. The last time Wall Street priced hardware volume over software margins was the cloud build-out of 2018–2019. Server OEMs surged. Two years later, the software companies owned the installed base and the margin pool had shifted entirely.
I'm not calling a top in Dell. A $27 billion guide raise and $24.4 billion in AI orders aren't the stuff of reversals. But the rate of margin compression — 430bp in a single quarter — sets a clock. At this pace, ISG margins hit single digits by Q3. At that point, Dell is genuinely a logistics company and the multiple contracts accordingly.
Falsifiable position: CRM outperforms MRVL over the next 30 days (by June 29). CRM at 13.6x with $3.88 EPS and $1.2B Agentforce ARR is the better risk-reward than MRVL at 64x with a penny beat. The market is mispricing the software layer because it's still in the volume phase of the AI cycle. When the cycle matures — and it always does — margin quality re-rates first.
Dell Technologies (DELL) Q1 FY2027. Reported May 28, 2026 AMC. Non-GAAP EPS $4.86 vs $2.93 est (+65.9%). Revenue $43.8B (+88% YoY) vs $35.5B est. ISG revenue $29.0B (+181% YoY). AI server revenue $16.1B (+757% YoY). ISG OM 10.5% (↓430bp QoQ). Gross margin 17.8% (↓330bp YoY). FY27 guide raised to $165–169B (+$27B). Stock $317 close → $441 AH (+39%).
Salesforce (CRM) Q1 FY2027. Reported May 27, 2026 AMC. Non-GAAP EPS $3.88 vs $3.12 est (+24.4%). Revenue $11.13B (+13% YoY). Non-GAAP OM 34.8%. FCF $6.6B. Agentforce ARR $1.2B (+205% YoY). cRPO $33.6B (+14%). $25B ASR. Q2 guide $11.27–11.35B (vs $11.36B est). FY27 guide $45.9–46.2B. Stock ~$176, flat.
Marvell Technology (MRVL) Q1 FY2027. Reported May 27, 2026 AMC. Non-GAAP EPS $0.80 vs $0.79 est (+1.3%). Revenue $2.418B (+28% YoY). Data center 76% of rev. FY28 guide raised to $16.5B. Custom silicon >$10B by FY29. Stock ~$200, −4.6%.