Key points
- Nvidia requested that Samsung, SK Hynix, and Micron produce base dies for its custom NVHBM memory using TSMC's 3-nanometer process.
- Samsung Electronics agreed to the request, marking its first-ever use of TSMC's manufacturing technology for this purpose.
- The arrangement involves producing base dies for Nvidia's high-bandwidth memory, a critical component for advanced artificial intelligence hardware.
Samsung Electronics has agreed to manufacture base dies for Nvidia’s custom high-bandwidth memory using TSMC’s 3-nanometer process. This agreement marks the first time Samsung has utilized TSMC’s foundry technology for production. The move follows a specific request from Nvidia to its key memory partners, including SK Hynix and Micron, to adopt this advanced manufacturing node for their upcoming components.
How it unfolded
- Nvidia approached Samsung, SK Hynix, and Micron with a request to manufacture base dies for its custom NVHBM memory using TSMC's 3nm process.
- Samsung Electronics agreed to the proposal, committing to produce the base dies using TSMC's advanced 3-nanometer manufacturing technology.
- This agreement represents Samsung's first utilization of TSMC's foundry services, signaling a significant shift in supply chain dynamics for high-end memory production.
Who is affected
Nvidia is the primary beneficiary of this arrangement, as it secures access to advanced manufacturing capabilities for its custom NVHBM memory. These memory modules are essential for powering high-performance artificial intelligence systems and data center infrastructure. By leveraging TSMC’s 3nm process, Nvidia aims to enhance the performance and efficiency of its next-generation AI hardware, potentially gaining a competitive edge in the rapidly evolving semiconductor market.
Samsung Electronics is significantly impacted by this deal, as it enters a new partnership with TSMC. Historically, Samsung has relied on its own internal foundry capabilities for memory production. This agreement marks a strategic shift, indicating that Samsung is willing to outsource certain critical manufacturing steps to competitors when necessary. This move may influence future supply chain strategies and competitive dynamics within the semiconductor industry.
SK Hynix and Micron were also approached by Nvidia for this manufacturing arrangement. While the source material does not specify their responses, their inclusion in the initial request highlights the competitive pressure all major memory manufacturers face. These companies must decide whether to adopt TSMC’s process or continue with their existing manufacturing methods, a decision that could affect their market position and technological leadership.
What happens next
The immediate next step involves Samsung beginning the production of base dies for Nvidia’s NVHBM using TSMC’s 3nm process. This production phase will test the integration of Samsung’s memory expertise with TSMC’s advanced logic manufacturing capabilities. Success in this venture could pave the way for further collaborations between memory manufacturers and foundry specialists, reshaping the industry landscape.
Industry observers will likely monitor the performance and yield rates of these newly manufactured base dies. If the partnership proves successful, it could encourage other memory manufacturers to consider similar arrangements with TSMC. This trend could accelerate the adoption of advanced process nodes across the semiconductor sector, driving innovation and competition in high-performance computing and AI applications.
The background in plain language
High-bandwidth memory (HBM) is a specialized type of memory designed to deliver massive amounts of data to processors quickly. It is crucial for artificial intelligence tasks that require rapid processing of large datasets. Base dies are the foundational layers of these memory stacks. Using a 3-nanometer process means the transistors on these chips are incredibly small, allowing for more components in a smaller space. This results in faster speeds and lower power consumption, which are vital for modern AI systems.
What to do and how to stay safe: Samsung
- Monitor official announcements from Nvidia and Samsung regarding the performance and availability of the new NVHBM modules.
- Evaluate your current hardware infrastructure to determine if upgrading to next-generation AI accelerators is necessary for your workload.
- Stay informed about supply chain developments in the semiconductor industry, as shifts in manufacturing partnerships can impact product availability and pricing.
- Consult with technology vendors to understand how advancements in memory technology may affect your long-term computing strategy.
Step-by-step guide: Mobile Device Security for Small Business: Control Without Cost
General security guidance from the Circuit Observer newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why is Samsung using TSMC's process for this project?
Samsung agreed to Nvidia's request to use TSMC's 3nm process for manufacturing base dies, marking its first use of TSMC's foundry technology.
Which other companies were asked to make these base dies?
Nvidia also asked SK Hynix and Micron to produce base dies for its custom NVHBM memory on TSMC's 3nm process.
What kind of memory is being produced?
The base dies are for Nvidia's custom NVHBM, which is a type of high-bandwidth memory used in advanced computing systems.
