Nomura Quintupled SK Hynix's Target in 6 Months — A Tailwind for MU and SNDK

By Edgen Research | 2026-05-18 Tickers: $MU, $SNDK | Related (landscape): SK Hynix (000660.KS), Samsung (005930.KS), Kioxia (285A.T)

On May 17, Nomura raised SK Hynix's price target to ₩4 million won — roughly five times the level it set just six months ago. The size of that revision matters less than what Nomura wrote underneath it: AI memory demand is entering a "new regime" that breaks the historical cycle model. We're updating our #66 AI hardware basket map on the back of it — MU and SNDK move from Hold to Buy.

Key Takeaways

• Nomura raised SK Hynix PT to ₩4M (current ₩1.82M, +120% upside) on May 17, 2026 — 5x from ~₩840K in November 2025, four sequential raises in six months.
• Backing Nomura's call: AI memory demand projected to grow ~1,000x over 5 years, supply growth capped at 5-6x (~30% CAGR). Hyperscalers signing 3-5 year LTAs with prepayments locking in pricing.
• Goldman lifted MU's 2026E EPS by 133% ($21.01 → $49.01) in a single December 17 note, with FY2Q26 guidance beating Street consensus by 29% on revenue and 69% on EPS.
• Nomura's December 24 macro note forecast the memory market growing 4.6x ($103B in 2024 → $481B by 2027F). Six months later, every PT raise across Wall Street has tracked that trajectory.
• We upgrade MU and SNDK from #66's "Hold, don't add" to "Buy" — both sit inside the same structural channel Nomura, Goldman, and consensus are now repricing.

Nomura raised SK Hynix's target 5x in six months

The trajectory is the story, not the endpoint.

Date	Nomura SK Hynix PT (KRW)	Move	Source
Nov 2025	~₩840K	from ~₩540K (+56%)	Nomura Korea Technology
Apr 24, 2026	₩2.34M	from ₩1.93M (+21%)	Nomura Asia Equity Research
May 17, 2026	₩4.0M	from ₩2.34M (+71%)	Nomura Global Memory: Real AI Boom Is Here

Four sequential raises. Each one looked aggressive when published. Each one was conservative within weeks. The current PT implies +119.9% upside from the May 15 close of ₩1.82M — Nomura's bull case explicitly recalibrating to a "new regime," not extrapolating from the prior cycle.

I'd argue the May 17 raise matters more than the four that came before it. The earlier raises tracked rising EPS estimates inside the same cyclical model. This one reframes the model itself — what Nomura calls "memory stock re-rating driven by exponential demand growth from AI."

Nomura's macro thesis: 1,000x demand vs 5-6x supply

Nomura's December 24 Global Memory note projected the total memory market growing 4.6x from 2024 to 2027:

The May 15 follow-up Real AI Boom Is Here sharpens the underlying math. From the report:

"Memory demand could rise by several thousand-fold over the next five years... In contrast, we believe industry supply growth is likely to remain constrained to roughly 5-6x over the same period (CAGR of c.30%), raising serious questions around whether structural undersupply can realistically be resolved."

The numbers are extreme on purpose. Nomura is arguing the entire memory-cycle mental model — supply catches up, ASPs collapse, exit before the roll — is now wrong. The variables that historically governed the cycle (consumer inventory swings, fab build cycles) are no longer the binding inputs. AI token consumption is. And AI token consumption is structurally exponential.

Nomura's view on the LTA framework — the second pillar of the structural reframe — is equally direct:

"Three-to-five years of minimum contract periods, prepayments, and capex support commitments appear increasingly common, making cancellation difficult and enhancing binding aspects of contracts, effectively guaranteeing near-current level of profitability."

This isn't a 2024 spot-market cycle. It's an infrastructure-grade procurement cycle, with contractual stability that didn't exist in prior memory upcycles.

Goldman saw the math in December — and is still Neutral

Six months before Nomura's May reframe, Goldman Sachs published a December 17, 2025 note on Micron with quietly extraordinary numbers:

The headline numbers from the Goldman note:

Metric	Goldman Old	Goldman New	Revision
2026E EPS	$21.01	$49.01	+133%
2027E EPS	$23.81	$38.47	+62%
Price target	$205	$235	+15%
MU FY2Q26 revenue guide midpoint	(Street: $14.46B)	$18.7B	+29% above Street
MU FY2Q26 EPS guide midpoint	(Street: $4.97)	$8.42	+69% above Street

Goldman raised its 2026 EPS estimate by 133% in a single revision — the kind of revision that usually only happens when a model has been structurally below reality for quarters. And Goldman still kept MU at Neutral, citing potential pricing retracement from Samsung HBM qualifications.

That conservatism, in retrospect, is the tell. MU has run ~140% above Goldman's December PT in the five months since. Bullish sell-side wasn't bullish enough.

Goldman also flagged what we'd call the LTA evidence directly. From the report:

"Micron has already completed its volume negotiations with customers for 2026... The company expects its supply of HBM products to continue to ramp throughout the year, with production volume shipments of HBM4 beginning in 2QCY26."

And on the HBM market structure:

"From an HBM baseline TAM of $35B in 2025, the industry can grow revenue at a 40% CAGR to over $100B by 2028."

Goldman's $35B → $100B HBM TAM forecast is the sub-segment math sitting underneath Nomura's $103B → $481B total memory market forecast. They're two views of the same structural channel.

US-equity reads: MU and SNDK move to Buy

Micron (MU) — the only US HBM supplier. Multiple sell-side desks have now caught up to the reality Goldman flagged in December: Morgan Stanley's December 18 note had MU's EPS guidance "beating consensus by 75%." UBS in early March projected next-year EPS could approach $85. JPMorgan named MU a top semiconductor pick in its 2026 outlook citing "supply tightness persisting beyond 2026."

The structural channel Nomura describes is the same one MU sits inside — HBM allocation, DRAM contract pricing, LTA-locked volume. The argument that #66 made — "this is already in the price" — assumed sell-side was modeling the cycle accurately in May. The evidence from Nomura's May 15 reframe is that even bullish sell-side has been chronically under-modeling. MU's structural channel hasn't been re-rated yet at the multiple Nomura is implying for SK Hynix.

SanDisk (SNDK) — the only US NAND pure-play. Kioxia's May 15 earnings confirmed the NAND side independently: Q1 FY2027 net profit guide of ¥869B (up 48x YoY), Q4 FY2026 operating margin 60%. BofA in early January raised SNDK PT to $390 citing "NAND's strategic position in AI inference rising." SNDK has since run materially past that number — same chronic-under-modeling pattern.

Our Micron vs SanDisk memory comparison covers the underlying pricing-power mechanism. The piece this article revises — #66 — placed both names in "Hold, don't add" on May 15. We move both to Buy.

The trigger is not any single PT number. It's three independent reads — Nomura on the macro, Goldman on MU specifically, Kioxia's actual print on NAND — pointing at the same structural reframe inside the same week.

3 signals that would put us back on Hold

Three signals — any one in motion over the next two quarters — would pull MU and SNDK back to Hold:

Risk	What to watch	Why it would break the thesis
HBM supply discipline breaks	Samsung HBM4 qualification accelerates beyond expected ramp	Adds 30%+ HBM supply faster than demand absorbs → ASP retracement, Goldman's stated concern
Hyperscaler capex slows	CSP capex guidance cut in 2H 2026 earnings	Removes the AI-spending channel that funds the structural demand thesis
CXMT (China DRAM) materially gains DRAM share	China-foundry DRAM bit growth >15% in 2027	Re-introduces commodity-DRAM oversupply dynamics outside the LTA framework

The fundamental data does not currently support any of these breaks. HBM ASPs are rising sequentially into Q4 2026 per Citi. CSP capex guidance is rising, not falling, into mid-2026. CXMT has yet to qualify any tier-1 hyperscaler customer at scale. But all three are real, and any one of them in motion in the next two quarters would warrant pulling the upgrade.

This article is research and education, not personalized investment advice. Edgen Research is the institutional byline used for cross-coverage synthesis pieces. Primary sources cited: Nomura Global Memory ("Real AI Boom Is Here," May 15, 2026; "Stronger Memory Chip Prices Boost Earnings," December 24, 2025); Goldman Sachs Micron Technology coverage (Dec 17, 2025 and follow-ups); supporting sell-side notes from Morgan Stanley, UBS, JPMorgan, Bernstein, Barclays, BofA Securities, Citi. Verify against current data before acting. Edgen and contributors may hold positions in the securities discussed.