Samsung Reportedly Adopts TSMC 3nm Solution for Nvidia HBM: Will SK Hynix and Micron Follow?

TradingKey - On October 18 ET, reports emerged of new supply chain arrangements for Nvidia's (NVDA) next-generation customized high-bandwidth memory (HBM).

According to South Korea's Seoul Economic Daily, Samsung Electronics may use TSMC's (TSM) 3nm process to produce base dies for Nvidia's customized HBM. Reuters subsequently picked up the report.

The report stated that Nvidia has also made similar requests to Samsung, SK Hynix (SKHY), and Micron Technology (MU), potentially expecting the three suppliers to use TSMC's process to manufacture base dies for NVHBM and next-generation DRAM processes for core dies.

Currently, Nvidia, Samsung, SK Hynix, Micron Technology, and TSMC have not officially confirmed the specific manufacturing arrangements, nor have order volumes, supply ratios, or mass production timelines been disclosed.

How Does NVHBM Differ From Standard HBM?

Nvidia has announced its NVLink Fusion platform for semi-custom AI infrastructure. Reuters, citing South Korean media reports, said Nvidia unveiled a custom HBM solution, NVHBM, in August.

Unlike standard HBM, NVHBM reportedly integrates Nvidia's custom memory controller into the base die. The base die typically handles logic control, interface connectivity, and memory management, while the core die primarily handles data storage.

This design is expected to allow Nvidia to unify product interfaces and validation standards across different suppliers, reducing product adaptation work and enhancing the feasibility of multi-vendor sourcing.

According to reports, NVHBM's design goals also include boosting memory bandwidth, reducing power consumption, and improving XPU space utilization. The ultimate results will still depend on chip design, manufacturing processes, packaging, yield rates, and customer qualification.

Why Samsung Might Adopt TSMC's Process Technology

If the reports are true, Samsung will use TSMC's 3-nanometer process to manufacture the base die for a specific NV HBM project, rather than relying entirely on its own foundry capacity. This may be related to Nvidia's desire to standardize base die solutions and streamline the supplier onboarding process.

This arrangement involves only specific customized projects and does not represent a complete shift in Samsung's production model. Reports state that Samsung still plans to retain its in-house manufacturing system for standard HBM and customized products for other customers, though this information has not been officially confirmed by Samsung.

For Samsung, adopting TSMC's process may help meet Nvidia's technical requirements and secure orders for next-generation custom HBM. However, whether the solution translates into actual supply still requires design verification, process qualification, yield improvement, and mass production ramp-up.

Will SK Hynix and Micron Adopt the Same Approach?

Currently, there is no public information indicating that SK Hynix and Micron Technology have accepted TSMC's base die manufacturing plan. Reports only state that Nvidia has made similar requests to three major HBM suppliers.

Even if the three companies adopt a relatively standardized base die design, differences among suppliers will still be reflected in core die process technology, HBM stacking, advanced packaging, yield rates, production capacity, and delivery stability.

For SK Hynix, NVHBM may bring new customized memory orders, but it also requires its products to conform to Nvidia's design standards. Whether Micron Technology can secure related orders will depend on customer qualification, capacity allocation, product performance, and pricing.

At present, it remains unclear whether the two companies will adopt the same manufacturing scheme as Samsung.

What Does It Mean for NVDA, Samsung and TSMC?

For Nvidia, a unified NVHBM solution could expand qualified supply sources, reduce dependence on a single memory vendor, and drive the co-design of custom HBM and XPU platforms. Whether this translates into actual shipments remains dependent on supplier qualification, order allocation, and mass production progress.

For Samsung, SK Hynix, and Micron, NVHBM presents potential high-end memory order opportunities while raising requirements for product specifications, manufacturing yields, and delivery capabilities.

For TSMC, if its 3nm process is ultimately adopted for the NVHBM Base Die, the application scope of advanced logic processes in AI memory could expand. However, there is currently no evidence indicating that TSMC has secured definitive orders, and the associated revenue contribution cannot be quantified at present.

Going forward, key areas to watch include whether the three memory makers confirm supply solutions, the progress of customer qualification and mass production for NVHBM, and whether Nvidia discloses specific suppliers, order volumes, and supply allocations.