SK Hynix Doubles Down on AI and Manufacturing with Major Strategic Push

South Korean semiconductor leader SK Hynix is launching a significant two-pronged offensive, combining a massive capital expenditure plan with a strategic initiative to shape the future of AI hardware. The moves underscore the company’s ambition to solidify its position in the high-stakes memory market.

Accelerated Fab Investment and National Ambitions

A central pillar of the strategy is a substantial new capital injection into its manufacturing footprint. The company has committed an additional 21.6 trillion Korean Won to a fabrication plant currently under construction in the Yongin cluster. This new funding phase is scheduled to run from 1 March 2026 to 31 December 2030.

When combined with an earlier tranche of 9.4 trillion Won approved in July 2024, the total investment at the Yongin site will reach 31 trillion Won. The fresh capital is earmarked for the construction of five additional clean rooms, bringing the facility’s total to six. Notably, the project timeline is being accelerated, with the opening of the first clean room moved forward from May 2027 to February 2027. The long-term vision for the location includes building four separate semiconductor fabrication plants.

This expansion aligns with broader national objectives. South Korea has publicly stated its goal to capture 10 percent of the global chip market. The initiative receives support from a separate state-backed program, announced in 2024, which allocates 26 trillion Won to enhance regional infrastructure, including water and power supply.

Pioneering a New Memory Standard for AI Inference

Concurrently, SK Hynix is positioning itself at the forefront of AI hardware development through standardization. Together with Sandisk, the company held a consortium kick-off event in California on 25 February 2026 to promote a global standard for High Bandwidth Flash (HBF).

Should investors sell immediately? Or is it worth buying SK Hynix?

The proposed HBF technology is designed as an additional memory layer situated between High Bandwidth Memory (HBM) and SSDs. This architectural approach aims to boost energy efficiency and scalability specifically for AI inference workloads—the phase where trained models are deployed and executed, rather than during initial training. A key objective is lowering total cost of ownership. The standardization effort will formally proceed under a dedicated working group within the Open Compute Project.

A hardware roadmap is already in place, targeting first memory samples in the second half of 2026 and initial inference devices for sampling in early 2027. The company has cited 512 GB modules as part of the development, a capacity far exceeding the 64 GB limit of current-generation hardware.

Advancing Core Technology with LPDDR6

Beyond new standards, SK Hynix is also pushing forward in its core memory business. At the ISSCC 2026 conference, the manufacturer presented research on LPDDR6 architecture, which promises to reduce power consumption by 20% while increasing bandwidth by 50%. Target markets for this technology include data centers, edge devices, and automotive systems.

Investor sentiment appears strongly positive, reflecting the market’s focus on capacity expansion and next-generation AI memory. The company’s shares recently hit a new 52-week high of 1,061,000 KRW.

Looking ahead, the new investment period begins in early March, while the next tangible technology milestones are set for the second half of 2026 with the arrival of the first High Bandwidth Flash samples.