Samsung +28%, SK Hynix +60%: Why Chip Giants Are Diverging

Samsung +28%, SK Hynix +60%: Why Korea's Chip Giants Are Diverging — Nomura Sees 100%+ Upside

Two semiconductor giants. One AI memory supercycle. And yet their stock charts couldn't look more different. Over the past month, Samsung Electronics has climbed 28.2% while SK Hynix has rocketed 60.12% — more than double its larger rival's gain. Both make memory chips. Both are riding the same HBM (High Bandwidth Memory) wave. So what gives? A union strike at Samsung, a historic rift between retail and foreign investors, and a bold call from Nomura that's reshaping how the market values these names. Here's the breakdown.

The Great Divergence

The numbers are stark. Samsung Electronics closed the month up 28.2%. Not bad. But SK Hynix? Up 60.12% over the same window. That's the kind of gap you don't see between two companies in the same industry unless something structural is going on.

The simplest explanation is HBM. SK Hynix is effectively a pure-play on high-bandwidth memory, the specialized DRAM that sits next to AI accelerators. It supplies the bulk of HBM3E to Nvidia and has a commanding lead in the next-generation HBM4 ramp. Samsung, by contrast, is a sprawling conglomerate — memory, foundry, smartphones, appliances. When the AI trade is on, SK Hynix benefits more directly. When things get choppy, Samsung's diversification cushions the blow.

And choppy they got. On May 15, the KOSPI dropped 6.12% in a single session. Samsung fell 8.61%. SK Hynix shed 7.66%. It was a brutal day, but the selling wasn't driven by fundamentals — it was margin-call panic, which we'll get to.

Then there's the labor overhang. Samsung's union has called an 18-day total strike starting May 21. That's not a symbolic walkout — it's a production risk that foreign investors are taking seriously. SK Hynix has no equivalent headline hanging over it. One stock gets the clean AI narrative. The other gets the AI narrative plus strike uncertainty. The gap starts making sense.

Nomura's Structural Rerating

On May 16, Nomura dropped a research note that got everyone's attention. They hiked their Samsung target from 340,000 won to 590,000 won. SK Hynix went from 2,340,000 won to 4,000,000 won. That's roughly 73.5% upside for Samsung and 70.9% for SK Hynix from current levels — and that's after the stocks have already run hard.

The more interesting part is how Nomura arrived at these numbers. Historically, memory stocks trade on price-to-book (PBR) because the industry is cyclical — you buy when book value is crushed, you sell when margins peak. Nomura is now arguing that's the wrong lens. They've switched to price-to-earnings (PER) valuation, treating semiconductor memory as a structural growth story rather than a boom-bust cycle.

The thesis rests on a 5-year AI-driven memory demand expansion. Training and running large language models requires enormous amounts of high-bandwidth memory. As models get bigger and inference moves to the edge, HBM demand compounds. Nomura sees this as a secular shift, not a temporary spike. If they're right, the entire sector gets repriced — and these targets might be conservative.

Notably that Nomura isn't alone. Several global investment banks have been revising their semiconductor outlooks upward, but the explicit switch from PBR to PER is what made this particular note land hard in Seoul trading rooms.

The Retail-Foreign Standoff

Underneath the price action, there's an ongoing tug-of-war that's reached historic proportions. Over the past month, individual Korean investors have bought a net 4.96 trillion won worth of Samsung Electronics shares. That's retail conviction on a massive scale — mom-and-pop investors betting that the AI story lifts Samsung through the strike noise.

Foreign investors have done the opposite. They've unloaded 10.01 trillion won of Samsung stock in the same period. That's more than double what retail bought. The foreigners are citing union risk, won currency exposure, and a preference for the cleaner SK Hynix play. When the May 15 crash hit, much of the forced selling came from margin calls.

Which brings us to the margin debt number. On May 13, Kiwoom Securities reported that margin loans hit 3.587 trillion won — an all-time record. When you combine record leverage with a sudden 6% index drop, you get forced liquidations. That's what amplified the May 15 selloff, and it's a risk that hasn't fully cleared. The strike starting May 21 could trigger another wave if production disruptions spook lenders.

One way to read this: retail is betting on the Nomura structural thesis, while foreign money is treating the near-term risks as a reason to reduce exposure. Someone's going to be right. The 10-trillion-won question is who.

The Bigger Picture

Strip away the daily price moves and the margin-call noise, and you're left with a genuine structural shift. High-bandwidth memory isn't a commodity cycle — it's a technology race. SK Hynix currently leads. Samsung has the resources to catch up. Both are essential to the AI supply chain, and both are trading below where Nomura thinks they should be if the PER framework is correct.

The union strike is a real near-term headwind for Samsung, but strikes end. The 5-year demand trajectory doesn't hinge on 18 days in May. What matters more is whether Nomura's re-rating thesis gains wider acceptance. If memory stocks start trading on earnings multiples rather than book value, the entire sector reprices — and the divergence between Samsung and SK Hynix may narrow, even if SK Hynix keeps the HBM crown.

Sources: Korea Exchange (KRX) market data, Nomura Securities research note (May 16, 2026), Kiwoom Securities margin data, Reuters, Bloomberg.

My Take

I think this divergence is going to widen before it narrows. Here's my concrete opinion: SK Hynix has first-mover advantage in HBM that Samsung won't close in the next 12 months. I've been tracking HBM market share data, and SK Hynix's dominance in HBM3e is structural at this point — Samsung's late entry means they're playing catch-up on both yield and customer qualification.

My actionable recommendation: I'd overweight SK Hynix over Samsung in a Korea semiconductor portfolio. The HBM moat is widening, not shrinking, and SK Hynix's valuation (trading at a premium to Samsung) is justified by the growth differential. But I'd watch Samsung carefully — they have the R&D firepower to close the gap, just not on a 6-month timeline. For now, the divergence trade is still on.