Samsung Electronics will use TSMC’s 3-nanometre manufacturing technology to build base layers for Nvidia’s next-generation high-bandwidth memory (HBM) chips. This partnership aims to standardize production for Nvidia’s AI processors. Investors should track how this collaboration impacts supply chain efficiency and memory chip output in the growing AI hardware market.
In a significant move for the global semiconductor supply chain, Samsung Electronics has decided to utilize TSMC’s advanced 3-nanometre manufacturing process for the base dies of Nvidia’s next-generation high-bandwidth memory (HBM) chips. This collaboration marks a notable shift, as Samsung, usually a competitor to TSMC in the chip manufacturing (foundry) business, is adopting TSMC’s technology to meet specific design requirements set by Nvidia.
Why Nvidia is Standardizing Memory Chips
Nvidia is driving this change to simplify and improve the performance of its AI hardware. The company is pushing for a design known as NVHBM, where it integrates custom memory controllers directly into the base layer of the memory chip. To ensure consistent performance across its entire AI infrastructure, Nvidia is requiring its primary memory suppliers—Samsung, SK Hynix, and Micron Technology—to adopt a standardized manufacturing approach for these foundational layers. By using TSMC’s high-precision 3-nanometre process, Nvidia aims to ensure that its memory chips perform reliably across different supplier production lines.
Strategic Implications for Samsung
This partnership represents a calculated trade-off for Samsung. While the company continues to maintain its own in-house production for many standard memory products and other clients, it is opting for TSMC’s foundry services specifically for the Nvidia project. This approach allows Samsung to remain a key supplier for Nvidia while navigating the intense technical requirements of the AI sector. For investors, this move highlights the massive demand for HBM, which is the high-speed memory essential for running powerful AI processors.
Risks and Monitorables
While this partnership aims to secure supply for Nvidia, it introduces a new layer of complexity. Relying on TSMC’s foundry capacity for critical components means that Samsung’s production schedule for these specific chips will be linked to TSMC’s availability. If demand for AI hardware continues to surge, any bottleneck at the foundry level could potentially impact the rollout of these memory chips. Additionally, investors should monitor the competitive landscape, specifically the market share battle between Samsung and SK Hynix in the HBM sector. SK Hynix has held a strong position in providing HBM for AI chips, and the market will watch to see if this new strategy helps Samsung close that gap.
The critical next step for stakeholders will be observing the execution timeline. The industry will track how quickly Samsung can integrate this new manufacturing workflow and whether it can scale production to meet Nvidia’s aggressive demand for AI-ready components.