2023 Latest Semiconductor News from March 26th to April 1st

Kioxia and Western Digital revealed their co-developed 8th generation BiCS 3D NAND flash device with 218 active layers. The new IC has a record-breaking 3200 MT/s interface speed, which will enable developers to build high-performance storage subsystems (such as the industry's fastest client SSD) using fewer 3D NAND chips. According to relevant sources, in order to achieve 3200 MT/s I/O, the two companies borrowed from YMTC's Xtacking technology.

The first 8th generation BiCS 3D TLC NAND device launched this week features 1Tb (128GB) capacity, a quad-plane architecture that maximizes internal parallelism and performance, and an interface speed of 3200 MT/s (which will provide peak sequential read/ write speed of 400 MB/s). Kioxia and WD unveiled world's fastest 3D NAND flash memory together.

On the evening of March 31, Beijing time, Netcom China issued the "Announcement on Initiating a Cyber Security Review of Micron's Products Sales in China".

The announcement pointed out that in order to ensure the security of the key information infrastructure supply chain, prevent network security risks caused by hidden product problems, and maintain national security, in accordance with the "National Security Law of the People's Republic of China" and "Network Security Law of the People's Republic of China", the Cyber Security Review Office follows the " Network Security Review Measures”, implements network security review of Micron products sold in China.

NVIDIA said cryptocurrencies have not "delivered anything useful to society" despite the massive sales of the company's powerful processors to the industry. Other uses of ChatGPT are more valuable than mining cryptocurrencies, said by Michael Kagan, the company’s CFO.

The first version of ChatGPT was trained on a supercomputer consisting of about 10,000 Nvidia graphics cards. "With ChatGPT, everyone can now create his own machine, and own program: you just tell it what to do, and it will do it. If it doesn't work the way you want, you can tell it' I wanted something different'," Kagan told.

Samsung Electronics said on Sunday that Samsung Executive Chairman Lee Jae-yong had visited a semiconductor component manufacturing base operated by a Samsung affiliate in China, his first visit to China in three years as the COVID-19 pandemic disrupted business travel, Yonhap News Agency reported. Lee inspected a multi-layer ceramic capacitor (MLCC) production line in Tianjin fab. And the last time Lee visited Samsung chip factory was in May 2020 in Xi'an, China.

According to semiconductor-digest, the global high-bandwidth memory market is expected to grow from US$292.7 million in 2022 to US$3,433.8 million by 2031, at a CAGR of 31.3% during the forecast period 2023-2031.

This growth can be attributed to growing demand for high-performance computing, artificial intelligence, and graphics processing applications that require advanced memory solutions. Compared to traditional DRAM memory solutions, HBM technology offers higher bandwidth, lower power consumption and higher performance, making it the first choice for these applications.

At NVIDIA GPU Technology Conference, Nvidia showed the real product of Grace, which represented its entry into the ARM CPU market. So why does Nvidia choose to develop its own ARM CPU at the moment?

The launch of Grace can be said to have been planned for a long time, and may even be earlier than Nvidia announced the acquisition of ARM. At the 2022 GTC conference after Nvidia announced the termination of the acquisition of ARM, Nvidia still insisted on launching this product and made a big splash at this year’s GTC. Moreover, just like the original acquisition promise, Nvidia did not get the so-called "advantage". The core used by each of their die is still the public version of ARM Neoverse V2, but the D2D interconnection uses its patented technology NVLink-C2C, which has a large bandwidth, and it can achieve excellent results of 900GB/s.

Given that servers and data centers are still dominated by x86, why should Nvidia, a company that has always carried out the x86 ecology, enter the ARM CPU at this time? The answer is naturally that the demand has already begun to rise, otherwise Nvidia would not be able to do business at a loss.

According to the Digitimes report, TSMC’s major customers, MediaTek and Apple, strengthened their efforts to cut orders in the second quarter, and Nvidia, which benefited from the ChatGPT boom, did not increase orders at TSMC in the second half of the year as expected.

Huang Renzhao, TSMC’s CFO, pointed out at the conference at the beginning of the year that due to the interference of inventory adjustment and sluggish consumption, the revenue in the first quarter of this year is expected to be approximately US$16.7-17.5 billion, a quarterly decrease of 14.2%. At present, the market professionals predict that TSMC’s revenue in the second quarter will continue to decline. In the second half of the year, despite the launch of new iPhones and the support of a 6% increase in OEM quotations throughout the year, due to conservative orders from other major customers, the annual revenue will increase slightly.

As for the price issue, the market has recently heard that TSMC may raise prices in the second half of the year, but IC design manufacturers pointed out that there is no change in the price in the third quarter, and the price in the fourth quarter will not be negotiated until April.

SK Hynix CEO said on Wednesday that the company will ask the United States to exempt it for another year from chip restrictions against China.

"Talks between the South Korean and U.S. governments should go well," and "From our side, we will also take time and adjust our management plan. When the one-year authorization expires, we will apply for an extension again." he said.

In October, 2022, SK Hynix said it had received a one-year authorization from the U.S. Department of Commerce to provide Chinese factories with the equipment needed for chip production without seeking additional licensing requirements.