What Makes Sammy Run?
IDC is out with early predictions for the semiconductor market for 2025, expecting a 16% increase in value this year to $779.8b. Inside of that number are expectations for 24% growth in the memory market, with DRAM expected to be up 30.2% to $118b, with HBM (High Bandwidth Memory) for AI the driver. As a contrast, they predict that the smartphone market will grow only 3.3% this year, primarily in 2H as inventory levels are worked down, but the datacenter market will see 23.9% growth, with that jumping to 33.6% if you include the cloud and enterprise markets.
As we have noted previously, Samsung, the leader in the memory market, has been unable to capitalize on that leadership and has been losing ground to competitors, SK Hynix (000660.KS) and Micron (MU) in this very competitive market. Samsung has had problems with heat and power consumption in its HBM3e 8 and 12-layer products that caused yield issues and qualification delays at Nvidia (NVDA), contributing to the recent weakness, and among the pressures that have lead to Samsung’s recent executive level parings. The head of Samsung’s memory group made a public apology for the shortfall in 4Q’24, leading one to believe that even more intense pressure continues for Samsung to right the problems in the semiconductor division, and it seems they are taking additional steps toward other issues that might be contributing to the problems with the company’s semiconductor production.
Samsung has been developing a more advanced memory product (D1b) that is based on a 10nm node. They have been producing D1b memory, which is used in its DDR5 memory products, on a 12nm node since late 2024, which itself was delayed by almost a year. This node change has raised expectations about the 10nm product, based on higher density and even higher speeds than the D1b 12nm product. However it seems that as it continued the development of the 10nm product, Samsung has taken the very unusual step of changing the design of the D1b 12nm product, which has been in production for nearly a year, to improve yield. This typically means that the production process is also changed, an expensive procedure, with equipment having been ordered in 4Q ’24 (unconfirmed) and production of the D1b memory with the new design scheduled for 2Q or 3Q this year.
This was not a light decision on Samsung’s part, likely based on the need to push ahead of SK Hynix and Micron, both of whom use D1b memory in their HBM product, while Samsung is still using D1a memory in its HBM. If Samsung is able to meet it goal of producing the new design of D1b memory quickly, without yield issues, they will remain competitive until the D1b 10nm development is completed. If they face problems with the new design and process, they could lose more ground to their competitors and the idea of a design change mid-stream will be viewed as a mistake. It’s a big bet, but also indicates the urgency being placed on fixing Samsung’s memory issues and allowing them to capitalize on the strength of the memory business this year, an absolute necessity if they are to remain the leader in the memory space.
As we have noted previously, Samsung, the leader in the memory market, has been unable to capitalize on that leadership and has been losing ground to competitors, SK Hynix (000660.KS) and Micron (MU) in this very competitive market. Samsung has had problems with heat and power consumption in its HBM3e 8 and 12-layer products that caused yield issues and qualification delays at Nvidia (NVDA), contributing to the recent weakness, and among the pressures that have lead to Samsung’s recent executive level parings. The head of Samsung’s memory group made a public apology for the shortfall in 4Q’24, leading one to believe that even more intense pressure continues for Samsung to right the problems in the semiconductor division, and it seems they are taking additional steps toward other issues that might be contributing to the problems with the company’s semiconductor production.
Samsung has been developing a more advanced memory product (D1b) that is based on a 10nm node. They have been producing D1b memory, which is used in its DDR5 memory products, on a 12nm node since late 2024, which itself was delayed by almost a year. This node change has raised expectations about the 10nm product, based on higher density and even higher speeds than the D1b 12nm product. However it seems that as it continued the development of the 10nm product, Samsung has taken the very unusual step of changing the design of the D1b 12nm product, which has been in production for nearly a year, to improve yield. This typically means that the production process is also changed, an expensive procedure, with equipment having been ordered in 4Q ’24 (unconfirmed) and production of the D1b memory with the new design scheduled for 2Q or 3Q this year.
This was not a light decision on Samsung’s part, likely based on the need to push ahead of SK Hynix and Micron, both of whom use D1b memory in their HBM product, while Samsung is still using D1a memory in its HBM. If Samsung is able to meet it goal of producing the new design of D1b memory quickly, without yield issues, they will remain competitive until the D1b 10nm development is completed. If they face problems with the new design and process, they could lose more ground to their competitors and the idea of a design change mid-stream will be viewed as a mistake. It’s a big bet, but also indicates the urgency being placed on fixing Samsung’s memory issues and allowing them to capitalize on the strength of the memory business this year, an absolute necessity if they are to remain the leader in the memory space.
RSS Feed