SK Hynix shares hit 20-year high on AI deal with AWS

Shares of South Korea’s SK Hynix hit their highest level in over 20 years on June 17.

The jump follows reports that parent company SK Group plans to build the country’s largest AI data center in Ulsan with Amazon Web Services.

SK Telecom and SK Broadband are leading the initiative, supported by other SK Group affiliates, including SK Hynix.

The stock has risen nearly 50% in 2025, including a 3% gain on June 17.

SK Hynix is a major supplier of DRAM and a leader in high bandwidth memory (HBM), used in AI servers. It supplies key clients like Nvidia.

🔗 Source: CNBC

🧠 Food for thought

1️⃣ HBM dominance reshapes competitive dynamics in memory market

SK Hynix’s market surge reflects a fundamental shift in the memory chip industry, where specialization in AI-specific components is now determining market leadership.

The company has captured an estimated 70% of the high bandwidth memory (HBM) market, the critical memory type used in AI servers, which helped it overtake Samsung in the overall DRAM market for the first time ever with a 36% share compared to Samsung’s 34% 1.

While the traditional DRAM market has remained sluggish, HBM demand is so robust that products are reportedly sold out through 2025 and possibly into 2026, creating a significant competitive advantage for SK Hynix 1.

This power shift is further emphasized by Samsung’s struggles, with its market cap reportedly sliding to a 9-year low as it faces challenges capitalizing on AI-driven demand, including reported difficulties selling HBM products directly to the Chinese market 2.

The investment in a new AWS data center partnership further strengthens SK Hynix’s strategic position in the AI supply chain, potentially accelerating its market leadership as data center build-outs continue.

2️⃣ Specialized AI chip demand creates winners and losers among semiconductor manufacturers

The remarkable contrast between SK Hynix’s 50% stock price surge and Samsung’s market challenges demonstrates how AI is reshaping competitive advantages across the semiconductor industry.

Demand for AI-specific memory is projected to continue growing substantially, with total generative AI compute demand potentially reaching 25×10^30 FLOPs by 2030 according to McKinsey analysis 3.