Goldman Sachs Substantially Raises DRAM, NAND, and HBM Price Forecasts
Claire Weston
Goldman Sachs sharply raised price targets on SK Hynix, Samsung, and Kioxia, arguing the memory upcycle will extend through 2028 — and current mid-single-digit P/E multiples do not reflect that call.
How much did the price targets move — and what upside does each imply?
SK Hynix target raised to KRW 3.5 million at 9× P/E, implying roughly 53% upside from the latest close. Rating: reiterated Buy.
Samsung Electronics target raised to KRW 480,000; preferred shares to KRW 360,000 — also at 9× P/E, implying roughly 60% upside. Rating: reiterated Buy.
Kioxia upgraded to Buy outright, with a new target of ¥93,000 at 7.8× P/E. The driver: Goldman now expects NAND prices to stay "higher for longer."
In plain terms = Goldman sees all three as significantly undervalued, with implied upside between 50% and 60%.
Why does Goldman think this cycle is structurally different?
First, the demand anchor has shifted. Memory demand is moving from PCs and smartphones to AI servers and agentic AI, making demand visibility far higher than in past cycles.
Second, supply is naturally constrained. Cleanroom capacity is limited, and HBM — high-bandwidth memory paired with AI chips — consumes 3–4× the wafer capacity of conventional DRAM, squeezing traditional supply further.
Third, customers are signing harder contracts. Long-term agreements (LTAs) now include prepayments, price floors, breach penalties, and quarterly volume commitments. This means → downstream demand is no longer just a forecast — it is a binding obligation.
How much has supply growth actually slowed?
Goldman estimates conventional DRAM capacity CAGR will fall from 12% in 2017–2018 to 7–8% in 2026–2030.
Conventional DRAM supply CAGR is projected to drop from 19% in 2017–2018 to 15% in 2026–2028.
This reflects a structural shift: HBM is consuming so much wafer capacity that traditional DRAM supply growth is being permanently compressed.
If the cycle really is longer, how would valuations re-rate?
Most memory stocks still trade at mid-single-digit P/E multiples. Goldman believes this pricing embeds the assumption that the cycle will peak soon.
But if memory companies can prove sustained earnings power, valuations could converge toward moat-protected growth stocks — which typically trade near 20× P/E.
Put simply = the market is pricing in a "cyclical discount," and Goldman is betting that discount is too steep.
Content is for reference only, not financial advice.