[News] SK hynix Reportedly Builds Silicon Valley HBM Team to Boost Customer Co-Design With U.S. Big Tech

SK hynix is reportedly strengthening its HBM co-design capabilities in Silicon Valley as it deepens collaboration with global Big Tech customers. According to NewsPim, sources say SK hynix America operates an HBM Architect Design Team in San Jose, California, working directly with major U.S. customers on base-die design for HBM and 3D-stacked DRAM. The team develops next-generation HBM architectures around customer requirements, going beyond local technical support to collaborate throughout product development.

Notably, NewsPim adds that SK hynix is expanding the team, offering annual salaries of US$150,000–US$260,000. The company is seeking experience in DRAM and HBM architecture, full-custom and digital co-design, power delivery network (PDN) analysis, and logic foundry design, with experience in 1x-nanometer DRAM and foundry nodes of 3nm or below preferred. Developing advanced, customer-tailored HBM architectures for future products is among the team’s core responsibilities, the report notes.

The San Jose location also supports this strategy. As NewsPim notes, NVIDIA and AMD are headquartered in nearby Santa Clara, while Broadcom has operations in San Jose. Positioning HBM engineers near these major AI chip design hubs allows SK hynix to work directly with customers on next-generation products and quickly relay their requirements to development teams in South Korea.

SK hynix is already deepening such collaboration with NVIDIA. Chosun Ilbo notes that the two companies are jointly developing next-generation memory for AI factories, expanding their partnership beyond supplying existing HBM toward developing advanced memory aligned with NVIDIA’s next-generation roadmap.

Customer Co-Design Reshapes HBM Competition

Customer co-design is expected to become even more important with HBM4 and future generations. As NewsPim notes, increasingly diverse AI accelerator architectures are driving different requirements for HBM capacity, bandwidth, and power, making early co-optimization of AI chips and memory increasingly important.

This is also reshaping HBM competition. Memory suppliers that participate early in customers’ next-generation AI chip development can incorporate specific requirements into their designs, while co-developed HBM may raise barriers for rivals seeking to enter the same supply chain later. As the report points out, such collaboration is becoming increasingly important for securing next-generation HBM orders.

The trend extends beyond SK hynix. Seoul Economic Daily notes that Samsung is recruiting base-die designers as customer-specific optimization for AI accelerator vendors becomes increasingly important with HBM4. The company is also hiring application engineers to evaluate HBM on AI accelerator systems and support key customer qualifications, with the aim of shortening the certification process.

Read more

• [News] SK hynix to Supply Memory for NVIDIA Vera CPU as Partnership Deepens; Samsung Meeting in Spotlight
• [News] Samsung, SK hynix 1H26 Chip Facility Investment Up 35%; NVIDIA Not Among Samsung’s Top Five Customers

(Photo credit: Samsung)