SK Group Courts US Tech Giants for AI Chip Alliance
South Korea's SK Group is actively courting Nvidia, Google, and Microsoft to form a new AI chip alliance. The move reflects the intense hyperscale demand for next-generation AI hardware. It also signals strategic maneuvering among global tech firms to secure semiconductor supply chains and innovation pipelines.
- SK Group Chairman Chey Tae-won personally met with the CEOs of Nvidia, Google, Microsoft, Meta, and Broadcom to move beyond a simple memory supplier relationship to one of joint product planning and co-design. - The alliance is centered on SK Hynix's High-Bandwidth Memory (HBM), a critical component for AI accelerators where the company holds a dominant market share, reportedly 62% in the second quarter of 2025. - Discussions with Nvidia involve securing the HBM4 supply for its next-generation "Vera Rubin" AI accelerator platform and collaborating on an "AI Factory" in Korea using over 50,000 Nvidia GPUs. - The talks aim to create custom-optimized memory for proprietary silicon, including HBM for Google's Tensor Processing Units (TPUs), Microsoft's "Maia 200" accelerator, and Meta's "MTIA" chips. - With Broadcom, SK Hynix plans to integrate its memory technology at the earliest design stages of the custom AI accelerators Broadcom builds for other large tech clients. - This push is part of a broader "AI Pyramid Strategy" by subsidiary SK Telecom, which includes developing its own Sapeon X330 inference chip and doubling its data center capacity by 2030. - The strategy also includes establishing a US-based investment arm, tentatively named "AI Co.," to fund innovative American AI companies and create synergies with SK's semiconductor manufacturing. - SK Group is already making substantial investments in the U.S., having previously announced a $22 billion plan for semiconductors, green energy, and bioscience.
