2nm GAA以及1.4nm皆在筹备中! 三星芯片宏图重磅出炉 向台积电发起最强挑战

The development of 2nm GAA and 1.4nm is in preparation! Samsung's chip grand plan has been launched, posing the strongest challenge to Taiwan Semiconductor.

Zhitong Finance ·  Jun 13 09:28

Samsung Electronics has released a series of new chip technology roadmaps that will be launched successively. The goal is to strengthen its HBM market share and attract top global AI chip companies to list Samsung as their preferred AI chip contract manufacturer.

It was reported that Samsung Electronics, the Korean storage chip giant, has released a series of new chip technology roadmaps that will be launched successively. The goal is to strengthen its HBM market share and attract top global AI chip companies to list Samsung as their preferred AI chip contract manufacturer, trying to launch the most powerful attack on TSMC who is known as the 'King of Global Chip Manufacturing' in the contract manufacturing field. Currently, the AI chip production capacity of chip giants such as Nvidia and AMD is almost fully concentrated in TSMC. The chip technology roadmap shows that Samsung plans to provide GAA advanced process technology on the upcoming 2nm process, and the preparation for its 1.4nm level chip process is progressing smoothly. The high performance and yield targets are expected to be mass-produced in 2027.

Although Samsung is currently the world's largest storage chip manufacturer, it has been trying to catch up with its rival TSMC in the chip contract manufacturing market. The chip contract manufacturing market mainly refers to the chip design enterprises (Fabless) that produce chips without wafer factories by foundries such as TSMC, such as high-performance AI chips designed by Nvidia, AMD, Broadcom, and Apple.

More specifically, Samsung belongs to the IDM model of chip companies, and enterprises in this model have the three core capabilities of chip design, manufacturing, and testing. They are also the highest technological threshold, the largest capital consumption, and the highest risk factors in the global chip industry chain., and Samsung and Intel are the world's top IDM chip companies.TSMC is mainly responsible for chip contract manufacturing, and belongs to the Foundry model in the chip industry chain.

On Wednesday in the Eastern United States, Samsung held its annual wafer contract manufacturing forum at its U.S. chip headquarters in San Jose, California, announcing its latest contract manufacturing and manufacturing roadmap, and outlining its many visions and prospects for the AI era.

It may be very difficult to challenge TSMC's leadership position in the contract manufacturing field in the short term

According to the latest data from market research firm TrendForce, in the first quarter of this year, Samsung's share of the chip contract manufacturing market unexpectedly slipped from 11.3% in the previous quarter to 11%, while TSMC's share of the chip contract manufacturing market rose from 61.2% in the same period to 61.7% higher than expected.

Under the extremely strong demand for chip components of high-performance computing systems for AI, the profitability of the Korean chip manufacturer is showing a positive recovery trend. The global trend of enterprises deploying AI has greatly boosted its main storage chip department and provided opportunities to win large-scale chip contract manufacturing contracts.

Samsung is the world's largest supplier of DRAM and NAND storage chips, and is also striving to become one of the HBM and the latest generation of HBM3E suppliers for Nvidia. In the field of DDR series storage chips (such as DDR4, DDR5) and SSD and other storage subdivisions, Samsung's market share is far ahead of other storage chip manufacturers. Unlike HBM, which is widely used in AI data centers, DDR series storage is mainly used in PC system memory to provide sufficient memory capacity and bandwidth, support multitasking processing and the processing of data sets of consumer electronics. LPDDR (Low Power DDR) series is used in smartphones.

However, Samsung must prove that its products are advanced and reliable enough to attract demanding major customers such as Nvidia to make larger chip contract manufacturing commitments. TSMC is currently still the most important chi p manufacturer for fabless chip design companies such as Apple, Nvidia, AMD, and Broadcom, especially for AI chips for data center servers that are being produced for Nvidia and AMD. It is considered the key hardware infrastructure driving large-scale artificial intelligence training/inference systems behind generative AI tools such as ChatGPT.

In addition, Samsung also faces chip contract manufacturing challenges from Intel, which is continuously opening large-scale chip contract manufacturing factories to accelerate its entry into the contract manufacturing field, trying to win chip-driven orders from chip companies such as Nvidia and AMD.

2nm GAA and 1.4nm advanced processes are actively preparing.

The progress of chip production technology is usually marked by the shrinking of transistor size, which helps to improve the core performance of electronic components. Competing for smaller sizes and mastering GAA technology may be key for Samsung to win substantial orders for AI processors, which are currently some of the highest performance and most expensive 4nm and 5nm advanced process chips, and are expected to concentrate on 3nm or even 2nm and below processes in the future.

Samsung's advanced process uses the so-called 'back power delivery network technology', which places power rails on the back of the silicon wafer. The company says that compared to the first generation 2nm advanced process technology, this technology improves power, performance, and area, while significantly reducing voltage drop area.

Samsung executives also emphasized that the company's comprehensive capabilities in logical chips, HBM storage systems, and chiplet advanced packaging will help it make rapid progress in winning substantial orders for outsourced manufacturing of AI-related chips.

Samsung predicts that by 2028, the number of customers for its AI chip-related products will increase by fivefold, and sales are expected to increase ninefold. The company announced several new production technologies and the latest layout for future AI-related chips, stating that this will help win more customers.

Samsung executives refused to comment on the latest progress in supplying Nvidia with the latest HBM storage system, or respond to media reports that Samsung has not yet obtained qualification to supply Nvidia with HBM.

Samsung also stepped up its promotion of its all-around gate-all-around (GAA) transistor chip manufacturing technology, which is a key technology for AI chips to move into 2nm and below processes. Samsung plans to begin mass production of the second generation 3nm process in the second half of this year, and provide GAA advanced process technology on the upcoming 2nm process. It is reported that Samsung was the first in the industry to begin mass production of 3nm chips based on GAA technology in 2022. In addition, the chip maker confirmed on the same day that preparations for its 1.4nm-level chip process are progressing smoothly, with high-performance and volume targets expected to be achieved in mass production by 2027.

The translation is provided by third-party software.

The above content is for informational or educational purposes only and does not constitute any investment advice related to Futu. Although we strive to ensure the truthfulness, accuracy, and originality of all such content, we cannot guarantee it.
    Write a comment