●Updates for HBM and Samsung Electronics
☆ We will summarize it in Osaka from afar. Some of the information below may be true, but please refer to it only for your information as it contains my personal judgment.
- Today, there was an article about the conditional approval from Samsung Electronics, citing the lack of supply of the HBM3e 8th tier by Nvidia. This seems to be a fact to some extent, and the HBM3e 8th tier, which is currently in the Nvidia H200, is in the Shortage situation. SK Hynix and Micron are among the suppliers of the HBM3e 8th tier to Nvidia, because Micron’s supply is still limited.
- Recently, there have been many negative and pessimistic opinions about Samsung Electronics. I indirectly conveyed my opinion on Samsung Electronics during an e-trend broadcast after the closing on Friday last week. I couldn’t explain the details, saying, “If you draw a line on Samsung Electronics’ price today and remember today as we check it at 1 month, 3 months, and 6 months, it will be of great help to invest in the future.”
- No one has received Nvidia’s 12th tier quality yet. SK Hynix has not passed the 12th tier quality yet. Personally, I expect SK Hynix’s 12th tier supply to start within the end of the year, and Micron’s supply to start sometime in the first quarter of 2025.
- What is important is that SK Hynix’s monopoly structure is likely to be resolved from the 12th stage of HBM3e, which is applied to Blackwell. Of course, I admit that it will supply more than the other two companies as the No. 1 company.
- It was from this third quarter that Hanmi Semiconductor’s sales were generated by microns, and we believe that the yield is coming up well from the 12th stage. Not long ago, an article said that Samsung Electronics’ HBM3e’s qualification was virtually over, but we are still conducting Nvidia’s HBM3e 12-layer qualification test. Recently, Nvidia seems to have started helping Samsung Electronics design HBM, indicating that it is highly willing to diversify its supply chain.
- We tend to think of HBM as a single product even if it’s the same HBM3, but there are dozens of types just for servers. Nvidia and AMD products are different. Samsung Electronics didn’t get HBM3e Qualification from Nvidia, but got Qualification from AMD.
- For the Nvidia B100, the HBM3e 8th tier will be the main product for the time being, and the 12th tier will have to be seen as the B200A. With the B100 being mass-produced, we don’t have to hurry to mass-produce the HBM3e 12th tier. Since the B200A is mass-produced in 25 years and 2Q, we have more time to supply the 12th tier than the market thinks. Therefore, I think concerns about oversupply of HBM in 2025 are excessive.
- With Nvidia currently accounting for more than 94% of global GPUs and AMD accounting for about 4% of the rest, it is correct to judge through Nvidia’s timeline, but for AMD, I think Samsung Electronics’ HBM3e 12th tier will be supplied from the end of the year. The logic that Samsung Electronics cannot do HBM3e in the first place is not correct.
- There are dozens of hundreds of different types of HBMs because they are customized differently by customer and application. Nvidia has a harder condition than AMD, so SK Hynix, which responded well first, is advantageous. It is true that SK Hynix’s technology is high, but it is not an overwhelming difference, and GPU companies such as Nvidia and AMD are pushing for customer diversification and prioritizing the transition to a consumer-oriented market.
- In September, Samsung Electronics reversed the previous process and started preparing for its new DRAM 4Q 25 years ago. Unlike SK Hynix and Micron, Samsung Electronics will proceed with process 1a. Since Samsung Electronics’ HBM yield was problematic in the entire DRAM process, it seems that it will change process 1a and move on to process 1b.
- Of course, there was a big problem with the original die design of process 1a, and this is a situation that affects process 1b, but the design change has also recently started, and we are trying to change it to the target of process 1d in 2027. Since the original die design of process 1a is now affected by the wrong part, and it usually requires about 3 years of R&D, we are trying to solve the fault even though it is late to the target in 2027. On the other hand, in process 1b, the technology migration to solve the problem of the entire process is strongly entering.
- The yield of HBM is actually large not only after-process but also after-process, and the yield may face a different phase due to unexpected issue. For example, there was a recent issue of seizure and search by DIT, and there is a possibility that SK Hynix will not be able to use the laser annealing equipment that is being used from process 1b. If the annealing equipment is not used, it is expected to affect the yield of HBM by about 10%. Therefore, yield changes and nothing is absolute.
- We absolutely judge Bonder’s technology in HBM manufacturing, but there are many factors that actually affect the yield. PCBAD (Plasma, CMP, Bonder, annealing, Dicing) and others greatly affect it, and previously, etching and deposition processes at wafer stage also affect it. Samsung Electronics recently started to review the equipment supply chain again.
- In a recent article, the fact that Samsung Electronics is focusing on HBM4 for a reversal means a change in its major definition, and it should be considered that it is targeting the fourth quarter of 25 and that there will be a big change in HBM4 on the road map.
- HBM4 is expected to change its major, and hybrid bonding will be adopted in the post-process, so looking for companies that benefit from it will be an important investment point in the semiconductor sector in 2025.
- Hybrid bonding requires plasma activation for low-temperature bonding. Cleanliness for intermolecular bonding and the process of flattening the surface are important. Accordingly, laser full cut, CMP, cleaning, and heat treatment balls