Abstract:
Samsung Electronics is adjusting the resource allocation of its advanced chip manufacturing business, shifting part of its 4-nanometer production capacity to high-bandwidth memory (HBM) base chips, and plans to use a more advanced 2-nanometer process to manufacture next-generation 5G modems. This adjustment will not only help Samsung increase the production capacity of advanced storage products, but may also create conditions for it to win Qualcomm's advanced chip foundry orders, giving Qualcomm the opportunity to reduce its sole reliance on TSMC in the future.
Samsung plans to shift its next-generation external 5G modem M2800 to 2nm process manufacturing. The chip is developed by Samsung's System LSI department and is expected to be completed in 2027. It aims to work with the Galaxy S28 series of mobile phones launched in 2028 and to work with the next-generation Exynos 2800 processor.
Previously, Samsung’s external modems mainly used the 4nm process. Shifting the M2800 to 2nm manufacturing means that Samsung needs to further improve the application of advanced processes in baseband communication chips, while also freeing up some 4nm production capacity for the manufacture of increasingly important HBM base chips.
HBM is a key component of artificial intelligence accelerators and high-performance computing systems. It provides higher memory bandwidth in a smaller space by vertically stacking multiple layers of DRAM chips. There is usually a base chip at the bottom of the entire stacked structure, which is responsible for control logic, data transmission interfaces, and connections with other chips. The design and manufacturing process of the base chip will directly affect the performance, power consumption and integration capabilities of HBM.
As the demand for artificial intelligence computing continues to grow, HBM’s market demand is expanding rapidly. Samsung not only needs to increase the production of DRAM core chips, but also must ensure that base chips can keep up with the production pace. Previously, Samsung mainly used its own 4-nanometer process to manufacture related substrate chips. Therefore, the allocation of 4-nanometer production capacity is becoming an important factor in the expansion of its advanced storage business.
Samsung’s shift of some 4nm resources to HBM manufacturing does not mean that all related products will immediately adopt the 2nm process. According to the information disclosed so far, Samsung still plans to use the 4nm process to produce the base chips of standard HBM products, while the next-generation M2800 modem will shift to the 2nm process. What the two adjustments have in common is that Samsung is re-planning the demand for advanced manufacturing capacity for different products.
This change is also related to NVIDIA’s customized HBM project. Previously, it was reported that Nvidia is cooperating with storage manufacturers such as Samsung, SK Hynix and Micron to develop customized HBM products, NVHBM, and requires relevant suppliers to use TSMC's 3nm process to manufacture base chips in order to unify the design and performance of key components.
However, Samsung has not completely given up using its own wafer foundry process to produce related base chips. Industry sources say that Samsung is verifying the feasibility of using its own 2-nanometer process to manufacture NVHBM substrate chips so that it can have an alternative when TSMC's advanced process capacity is tight. The relevant plans are still in the verification stage and cannot yet be regarded as confirmed large-scale mass production arrangements.
TSMC’s current 3-nanometer production capacity simultaneously serves multiple markets such as artificial intelligence accelerators, high-performance computing chips, and smartphone processors. As demand in these areas continues to grow, competition for advanced process capacity is becoming increasingly fierce. If Samsung can use its own 2nm process to produce substrate chips that meet the requirements, it will have the opportunity to provide additional manufacturing options for some customers.
At the same time, the capacity utilization and yield rate of Samsung's 2-nanometer process are also important factors affecting its future competitiveness. Analysis cited by Wccftech believes that if products such as Exynos 2700 and M2800 gradually adopt the 2-nanometer process, Samsung will have the opportunity to expand the actual application scale of this process. As capacity utilization increases and production experience accumulates, unit manufacturing costs may further decline.
This change may affect the foundry cooperation between Samsung and Qualcomm. For a long time, Qualcomm’s flagship Snapdragon processors have mainly relied on TSMC’s advanced manufacturing processes. TSMC's technological maturity and yield performance give it a competitive advantage in the field of high-performance mobile chips, but the cost of advanced manufacturing services is also relatively high.
If Samsung can improve yields, expand production capacity on the 2nm process, and provide competitive quotes, Qualcomm may consider adopting a dual-generation factory strategy in the future and hand over some products to Samsung for manufacturing, instead of concentrating the main production tasks on one TSMC company.
For Qualcomm, the value of the dual-factory strategy lies not only in potential cost savings, but also in supply chain flexibility. Advanced chip production is highly dependent on a small number of foundries. Once a certain manufacturer experiences capacity constraints, production interruptions or price changes, customers may be affected. Introducing a second qualified supplier can provide more options for product planning and capacity arrangement.
However, there is currently no evidence that Qualcomm has decided to hand over a certain flagship Snapdragon processor to Samsung for production using the 2nm process. Whether Samsung can obtain relevant orders still depends on actual yield, performance, power consumption, manufacturing costs, and Qualcomm's requirements for product reliability. Relevant reports suggest a possible development direction, rather than the confirmed results of commercial cooperation.
Samsung’s adjustment reflects that the demand for HBM brought about by artificial intelligence is changing the production capacity allocation of the entire semiconductor industry. For Samsung, HBM substrate chips are not only an integral part of storage products, but also an important growth opportunity for advanced wafer foundry business. By reallocating 4nm production capacity and promoting the application of 2nm processes to more products, Samsung hopes to enhance the competitiveness of its storage business and foundry business at the same time.
If this strategy makes progress, Samsung may not only expand HBM's supply capabilities in the future, but may also use the 2-nanometer process to win more high-performance computing and mobile chip orders. However, what really determines whether it can challenge TSMC is the yield, stability, mass production capabilities and customer verification results of the advanced process. For Qualcomm, whether Samsung can become a reliable source of second-generation manufacturing still needs to wait for subsequent product testing and business negotiations to give the answer.
Comments