Dell's Earnings Surprise and Korean Semiconductors: Pricing Power Sits With Components, Not Server Assembly

Fri, 29 May 2026 22:00:00 +0900

📚 Context for follow-up posts This is a follow-up to Marvell Q1 FY2027 and Korean Semiconductors. Where the Marvell post argued that “the AI bottleneck migrates from the GPU down to interconnect, substrate, and power,” this Dell post re-checks “who actually earns the money at that bottleneck” — through margin numbers. Related hubs: AI HBM Hub, Korea Semiconductor Value Chain Hub.

TL;DR

Dell reported FY2027 Q1 results on May 28, 2026 (U.S. time), and by any measure the numbers were a large earnings surprise. Revenue grew 88% in a single year, and adjusted EPS jumped 214%. The next day (May 29) the stock surged roughly 33%.

But buried in the same release was one uncomfortable number. No matter how many AI servers Dell sold, gross margin actually fell from 21.6% a year ago to 18.1%. Management’s own stated AI-server operating-margin target is mid-single-digit. In other words, Dell sells an enormous volume of product, but the margin on that product is thin.

That single line is the entire point of this post.

A company that assembles AI servers can post big revenue but thin margins. The real pricing power and the fat margin sit upstream — in memory and high-value components.

Dell itself said demand exceeds supply and that the primary constraint is memory. The moment the company building the servers says “what’s holding me back is memory,” it becomes clear where investors should look.

The Korea-translated priority order:

Priority	Korea Read-Through	Key Names	View
1	Memory pricing power (HBM, server DRAM, eSSD)	Samsung Electronics, SK Hynix	Most direct. SK Hynix is the best company but richly priced; Samsung has HBM4E / eSSD / foundry base-die catch-up optionality
2	FC-BGA, high-end MLCC, silicon capacitor	Samsung Electro-Mechanics	Direct second-order beneficiary, but P/E already in the 200s — “right direction, wrong price” risk
3	Datacenter power (second-order)	HD Hyundai Electric, LS Electric	Not a direct beneficiary of Dell’s print itself; a side-branch of the larger AI-power-demand trend
4	Avoid	Low-margin server assembly / set makers	The very spot Dell just illustrated: big revenue, thin margin

Dell itself had a strong quarter, but the source’s house view was HOLD, with a $475 12-month target. The discipline is not “AI servers are good → buy everything,” but to selectively look upstream, where margin and pricing power flow.

1. Dell’s Print: Revenue Explodes, Margin Compresses

First, the numbers as reported. (Dell)

Item	FY2027 Q1	Commentary
Revenue	$43.8B	+88% YoY; beat consensus (~$35.4–35.7B) by over $8B
GAAP diluted EPS	$5.24	+282% YoY
Non-GAAP diluted EPS	$4.86	+214% YoY; ~+$1.9 vs consensus (~$2.94–2.96)
ISG (servers, networking, storage)	$29.0B	+181% YoY
AI-optimized server revenue	$16.1B	+757% YoY
Traditional server/networking	$8.5B	+92%
Storage	$4.3B	+8%
CSG (PC)	$14.6B	+17% (Commercial $13.0B, +18%)
Operating cash flow	$4.1B	Adjusted FCF $3.165B, shareholder return $2.1B

Orders and guidance were strong too.

AI orders $24.4B; AI backlog $51.3B; over 5,000 AI customers.
Q2 revenue guidance $44–45B; Non-GAAP EPS midpoint $4.80.
FY2027 revenue $165–169B; Non-GAAP EPS midpoint $17.90.
FY2027 AI-server outlook ~$60B (+144%).

Up to this point it looks flawless. But the same release contains one decisive line.

Non-GAAP gross margin fell from 21.6% to 18.1%.

That means AI-server revenue grew more than sevenfold while the company-wide margin rate actually declined. Management put the AI-server operating-margin target at mid-single-digit. In other words, AI servers carry enormous revenue volume but a lower profitability profile than PCs or traditional servers.

2. Why the Margin Is Thin: The Value Sits in the Components

The reason is simple. The most expensive part of an AI server’s cost is not something Dell makes. It is the components: GPUs, HBM, server DRAM, eSSD, CPUs. Dell buys these, puts them in a chassis, connects the cabling, validates the system, and ships it to the customer.

As components get more expensive, revenue grows accordingly. But the margin on those expensive components is captured by the component makers. The assembler sees inflated revenue and thin margin. Dell’s print this quarter laid that structure bare.

The most important clue in management’s call was not the margin figure but the statement that demand exceeds supply, and the explanation that the primary constraint is memory. DRAM, NAND, CPU, HDD, and leading-node allocation were all cited as bottlenecks in the Q&A.

Translated into investor language:

It’s not that they can’t build the boxes — it’s that they can’t get enough of the key components. Whoever holds the scarce part sets the price. And the scarce part is memory.

If demand exceeds supply and the side that controls supply is memory, then pricing power clearly sits with memory. That is the starting point for the Korea read-through.

3. Korea Priority 1: Memory — Samsung Electronics and SK Hynix

The companies that sell the most of the very memory Dell called constrained are Samsung Electronics and SK Hynix.

Their Q1 results prove the structure directly.

Company	1Q26 Revenue	1Q26 Operating Profit	Note
Samsung Electronics	KRW 133.9T	KRW 57.2T	DS (semiconductor) division revenue KRW 81.7T, OP KRW 53.7T
SK Hynix	KRW 52.58T	KRW 37.61T	72% operating margin. HBM + high-capacity server DRAM + eSSD

SK Hynix’s 72% operating margin is no ordinary number. Set it next to Dell’s mid-single-digit AI-server operating-margin target, and you can see at a glance who captures the margin inside the same AI server. The side assembling the server is in the single digits; the side making the memory that goes inside it is in the 70s.

By name, the call differs.

SK Hynix is the strongest company in itself. It holds the core of AI memory: clean HBM exposure, high-capacity server DRAM, and eSSD. But as of May 29 its price is ~KRW 2,333,000, P/E ~22.1x, near its 52-week high. A great company, but the price is already elevated.
Samsung Electronics, as of May 29, is ~KRW 317,000, P/E ~25.4x, near its 52-week high (KRW 323,000). It is not as clean an HBM beta as SK Hynix, but it carries more catch-up optionality. It shipped the industry’s first HBM4E 12-high samples (in Q2), and eSSD plus a foundry-based base die also sit inside the same AI-memory stack.

For deeper discussion of Samsung Electronics, see Samsung Electronics Stock-Bonus Buyback Analysis and the Samsung Electronics Deep Dive.

The point: the moment Dell admitted that memory is the constraint, the pricing power of the two Korean companies holding that constraint structurally strengthens.

4. Korea Priority 2: High-Value Components — Samsung Electro-Mechanics

After memory, the next place pricing power flows is package substrates and power-stabilization components. In Korea, that seat is Samsung Electro-Mechanics.

Its Q1 results:

Item	1Q26	Note
Revenue	KRW 3.2091T
Operating profit	KRW 280.6B
Package-division revenue	KRW 725.0B	+45% YoY

The company directly cited strong demand for high-value MLCCs for AI servers and datacenters, and FC-BGA for AI/server applications. This dovetails precisely with the structure Dell revealed: as AI servers grow, demand for power-stabilization components and high-performance substrates around GPU/HBM packages grows alongside them.

For detailed analysis of Samsung Electro-Mechanics, see:

But here, price must be flagged. As of May 29, Samsung Electro-Mechanics is ~KRW 2,127,000, P/E ~227x. That is a very high number. It means the market has already priced in big expectations.

The direction is right. But the price is expensive.

Dell’s print did confirm that high-value components are positive, but Samsung Electro-Mechanics has already reflected much of that “good” in its price. So between memory (Priority 1) and Samsung Electro-Mechanics (Priority 2), there is a key difference: even the same beneficiary has a very different entry price.

5. Korea Priority 3: Datacenter Power (a Side Branch)

A lower-priority area is datacenter power. HD Hyundai Electric, as of May 29, is ~KRW 1,053,000, P/E ~48x.

The logic that more AI servers means more datacenter power demand, benefiting transformer and power-distribution makers, is sound. But this is not a direct result of Dell’s Q1 print — it is a side branch of the broader AI-power-demand trend. The constraint Dell explicitly named was not power; it was memory.

So power belongs in a second-order bucket: “related, but not the core read-through of this print.”

6. The Spot to Avoid: Low-Margin Assembly

The most important lesson from Dell’s print may lie less in what to buy than in what to avoid.

The fact that Dell sold more than seven times the AI servers and still saw its margin rate fall lays bare the structural limit of low-margin server assembly and set businesses. The revenue looks dazzling, but the money goes to the component makers.

So buying assembly/set-stage names indiscriminately just because they carry the “AI server beneficiary” label is dangerous. Even within the same AI-server revenue growth, the side with pricing power (memory, high-value components) and the side without it (assembly) are completely different investments.

7. Discipline: “AI Servers Are Good = Buy Everything” Is an Overreach

This print is clearly a strong positive. But the size of the positive and the investment call are separate things.

Three layers must be kept distinct.

Fact: Dell revenue +88%, EPS +214%, AI servers +757%, yet gross margin 21.6% → 18.1%. Memory is the primary constraint. These are numbers the company reported.
Inference: Margin and pricing power flow toward memory and high-value components rather than the assembler. In Korea the order is Samsung Electronics / SK Hynix → Samsung Electro-Mechanics → power. This is a judgment reasonably drawn from the facts.
Speculation: How much HBM4E a specific company supplies to Dell, and what share each maker holds in Dell’s servers, is not disclosed. This is not confirmed fact.

In particular, specific customer/share claims such as “Samsung Electronics supplies HBM4E to Dell” or “Samsung Electro-Mechanics holds X% of Dell server FC-BGA” are not confirmed in public disclosures. Such content should be classified as speculation or unverifiable, not fact.

Even a sound thesis ends as a theme unless it is validated in the numbers.

This post’s conclusion is not “AI servers are good, so buy all the related names.” It is: look upstream, where margin and pricing power flow — and select with the entry price in mind.

8. Next Checkpoints

Checkpoint	Strong Signal	Weak Signal
Dell AI-server margin	Mid-single-digit target achieved or raised	Margin falls further or misses target
Dell memory constraint	Memory shortage persists, sustaining pricing strength	Memory supply eases, prices soften
Samsung Electronics	HBM4E mass production / customer qualification, eSSD volume / pricing	Intensifying HBM competition, price declines
SK Hynix	HBM / server DRAM margins hold	2027 supply-overhang signals
Samsung Electro-Mechanics	Sustained package-revenue growth, confirmed margin	Growth decelerates while P/E stays high

Final Interpretation

Dell FY2027 Q1 is not a signal to “buy all AI servers” for Korean semiconductors. The more precise message is:

As AI-server revenue grows, margin and pricing power move up, away from the assembler and toward memory and high-value components.

From this vantage point, the most direct Korean names are memory — Samsung Electronics and SK Hynix. SK Hynix is the strongest company in itself but already priced high; Samsung Electronics carries HBM4E / eSSD / foundry base-die catch-up optionality. Samsung Electro-Mechanics is a direct second-order beneficiary through FC-BGA, high-end MLCC, and silicon capacitor, but at a P/E in the 200s it sits in a “right direction, wrong price” spot. Datacenter power is a side branch; low-margin assembly is the spot to avoid.

The conclusion is discipline. The stronger the positive, the more it matters not just what you buy, but at what price.

Fact / Inference / Speculation / Blocked

[Fact]

Dell FY2027 Q1 revenue was $43.8B (+88% YoY); Non-GAAP diluted EPS was $4.86 (+214%). (Dell)
ISG revenue $29.0B (+181%); AI-optimized server revenue $16.1B (+757%). (Dell)
Non-GAAP gross margin fell from 21.6% to 18.1%; AI-server operating-margin target is mid-single-digit. Memory (DRAM/NAND) was cited as the primary constraint. (Dell)
AI orders $24.4B; AI backlog $51.3B; FY2027 AI-server outlook ~$60B (+144%). (Dell)
Samsung Electronics 1Q26 revenue KRW 133.9T, OP KRW 57.2T; DS-division revenue KRW 81.7T, OP KRW 53.7T.
SK Hynix 1Q26 revenue KRW 52.58T, OP KRW 37.61T, 72% operating margin.
Samsung Electro-Mechanics 1Q26 revenue KRW 3.2091T, OP KRW 280.6B, package-division revenue KRW 725.0B (+45% YoY).

[Inference]

As AI-server revenue grows, margin and pricing power flow toward memory and high-value components rather than the assembler (Dell).
A reasonable Korea read-through priority is Samsung Electronics / SK Hynix (memory) → Samsung Electro-Mechanics (FC-BGA / MLCC / SiCap) → power (second-order).
SK Hynix is the strongest company in itself but carries a high valuation; Samsung Electronics has more catch-up optionality.

[Speculation]

Whether Samsung Electronics directly supplies HBM4E into Dell AI servers, and at what share, is not disclosed.
Samsung Electro-Mechanics’ share in Dell server FC-BGA / MLCC cannot be specified from public disclosures.

[Blocked]

Supplier identity and share by component within Dell’s AI-server BOM.
Samsung Electronics’ and Samsung Electro-Mechanics’ direct revenue exposure to Dell.
Real-time consensus EPS revision rates and NTM valuations for the Korean names.

Sources

Disclaimer: For research and information purposes only. Not personalized investment advice; this is public-information-based analysis. Names cited are for analytical illustration; readers should perform their own due diligence and consult licensed advisors before any investment decision.

델 on Korea Invest Insights