Barclays has sharply raised its price targets for two memory giants. The Frankfurt-listed SK Hynix target jumps from €1,100 to €2,300, while Samsung Electronics' London-listed target climbs from 425,000p to 730,000p (roughly £7,300). Both stocks retain "Overweight" ratings. This is Barclays' second upgrade in 2026, following increases in early May from €900 and $4,000 to €1,100 and $4,250 respectively.
HBM Sold Out Through 2027
Barclays argues the memory market's supply-demand imbalance shows "no near-term signs of improvement." Global memory supply is expected to grow just over 20% in both 2026 and 2027, but demand is accelerating faster, leaving the gap unclosed for two years. SK Hynix, Nvidia's key HBM supplier, has sold out its entire 2026 HBM capacity, with shortages extending into 2027. Customer demand exceeds planned output for the next three years.
In Q1 2026, SK Hynix posted revenue of 52.6 trillion won, up 198% year over year, with an operating margin of 72%. On May 27, its stock surged 11% in a single day, pushing market cap past $1 trillion — joining TSMC and Nvidia in the trillion-dollar club. Samsung reached that milestone earlier on May 6, becoming the second Asian semiconductor firm to do so after TSMC. Barclays expects Samsung's HBM revenue to triple in 2026 and raised its 2026 and 2027 revenue estimates by 8% and 17% respectively. By June 1, Samsung common stock market cap surpassed 2,000 trillion won, with the share price near 365,000 won.
Supercycle Accelerates
The memory industry is in a textbook supercycle, with Wall Street still chasing. Year to date, SK Hynix is up over 260% and Samsung over 200%, both driven by AI-fueled HBM demand. Goldman Sachs has also flagged that memory shortages could extend into 2028, with both Samsung and SK Hynix ramping up capacity. Meanwhile, downstream victims like GoPro have admitted to "material doubt" about operations and are exploring a sale or merger.
Barclays' target hikes coincide with Computex 2026, where AI hardware demand is being showcased en masse. HBM, the bottleneck for AI training chips, shows no near-term supply relief.