UBS Raises SK Hynix Target Price to KRW 3.2 Million
Alina Collins
UBS lifted its SK Hynix target from KRW 3.0 million to KRW 3.2 million, maintaining a buy rating; the bank forecasts 2027 operating profit at KRW 62.3 trillion — roughly 54% above consensus — arguing the stock still underprices a structural leap in earnings power.
Up 258% already — why does UBS still call it cheap?
At the current price, SK Hynix trades at just 2.76× 2027 P/B, implying a long-run ROE of roughly 31.7%.
This means → the market is pricing SK Hynix's long-term profitability at 31.7%, but UBS estimates the actual figure at 41.9% — that ten-point gap is the valuation upside UBS sees.
In plain terms = earnings expectations have risen faster than the share price; profit is outrunning price, so the stock looks cheaper than the headline gain suggests.
Three catalysts ahead: LTAs, HBM4, and buybacks?
Long-term agreements lock in volume and price. SK Hynix is signing 5-year-plus DDR5 and NAND contracts with hyperscale cloud customers, securing 60%–70% of planned shipments and pricing. This means → revenue visibility jumps sharply, partially smoothing the boom-bust cycles that have long weighed on memory valuations.
HBM4 mass production is imminent. Final design tweaks are largely done; large-scale shipments to Nvidia's Rubin platform are expected from Q2 2026. UBS projects HBM's share of DRAM revenue will climb from 15% in 2026 to 58% by 2030.
Post-ADR buyback programme. Market attention has shifted from "how much has the stock risen" to "how much cash will the company return" — a buyback launch would directly validate improving free cash flow.
How steep is the memory price ramp?
UBS forecasts SK Hynix DRAM ASP to rise 43% QoQ in Q2 2026; stripping out long-term-agreement effects, the DDR blended ASP jumps 67% QoQ.
Q3 and Q4 DRAM ASPs are projected to climb another 21% and 13% respectively; NAND blended ASPs rise 43%, 25%, and 10% across Q2–Q4.
This means → the price curve runs "steep early, gentle later" — the sharpest gains land in Q2, then taper but keep rising through year-end.
What is sustaining demand behind these price hikes?
Agentic AI — a new generation of AI that autonomously executes tasks — is pulling memory demand along multiple axes: AI servers need DDR5/LPDDR5, KV caches (high-speed temporary storage used during large-model inference) consume large amounts of DRAM, and storage drives require more NAND.
UBS projects 2027 DRAM bit consumption up 36% YoY, above 22% in 2026; NAND bit growth is also accelerating.
In plain terms = AI doesn't just eat compute — it is exceptionally hungry for memory. Larger models and more users mean more memory consumed, and that is the fundamental reason memory chips can keep repricing higher.
Which milestones should investors watch in H2?
LTA signing progress, HBM4 mass-shipment timing, and formal buyback announcement — whether these three lines deliver on schedule in H2 2026 is the core test of whether the market accepts this target-price upgrade.
This means → UBS's logic chain is clear but tightly linked: a delay in any single leg slows the entire re-rating.
Samsung may edge ahead in HBM bit share by 2027 at 41% vs. 39%; competitive pressure is another variable to track.
Content is for reference only, not financial advice.