Weekly News | Supply Chain Trends in Semiconductor Industry

Enterprise News(April 24

Apple’s M3 chips entering mass production, the first run of TSMC’s 3nm

According to tech website MacRumors, Apple’s M3 Chips will officially enter mass production in the second half of the year, using TSMC’s 3nm process, as rumor had it earlier. The M3 chip will be introduced in 2023 MacBook Air, iMac, and Mac mini, while the iPad series will wait until 2024.

As reported, the mass production of the M3 chip, codenamed Ibiza, will use TSMC’s 3nm process. The TSMC’s 3nm process that Apple‘s M3 is going to use is currently the most advanced chip node process. Compared to the 5nm process, it offers 60% higher logic density at the same speed while consuming 30% to 35% less power. Chips manufactured using TSMC’s 3nm process will yield better power consumption and performance while achieving higher transistor density, which helps to enhance Apple equipment’s running speed and efficiency and offer a better user experience.

However, according to leaked information, the FinFET process for producing 3nm chips faces extreme challenges. The A17 chips produced failed to reach the power consumption and heat generation standards, which pushed Apple to adjust A17’s performance goals. That means Apple has to drop A17 and M3 chips’ performance to reduce power consumption and heat generation. The overall performance target of A17 Bionic was decreased by 20% mainly because TSMC’s N3B node failed to meet the production goal.


On the one hand, Apple’s adoption of 3nm chips on its new MacBook Air and iPad Air/Pro demonstrates its continuous pursuit and efforts on technology innovation and its potential to push the entire industry to a higher process level. On the other hand, in the context of the reduced consumer electronics demand, the 3nm process’s usage is lower than expected, leading to some processes immature practices. To gain a time advantage in adopting TSMC’s N3 technology to achieve chips offering lower power consumption or higher performance under the same architecture and frequency, Apple inevitably has to pay a higher cost and the price of a lower performance expectation. In short, TSMC’s 3nm yields will be a direct constraint for A17 and M3 chips’ shipments.

Enterprise News(April 25

Chip architecture giant Arm getting into the chipmaking business

Arm recently announced that it had reached a cooperation agreement with Intel Foundry Services (IFS). Arm’s interested person in charge said the company had been cooperating with all wafer foundries, including TSMC, Samsung Foundry, and IFS, to facilitate the realization of the Arm architecture-based computing. Arm wants to build its own chips. As reported, Arm started to work on its chips six months ago. This chipmaking project is “more advanced” than ever, and a new “solutions engineering” team has been built to lead the new prototype chips’ development.

In fact, Arm’s chipmaking move has a lot to do with the IPO plan of its parent company Softbank, which acquired Arm for US$32 billion in 2016, and has been in poor financial health for a long time since. Although the technology licensing business model has been very profitable for Arm, it has also limited its profitability ceiling, lowering its maximum profits. Arm charges customers according to a certain percentage of the chips value. However, even if the proportion of chip licensing fees is 1% higher or more, it will not significantly increase the company’s overall revenue.


Changing Arm’s business model while increasing its research, development, and innovation investment might be a good strategy to enhance its market valuation further. Arm can showcase its design capabilities and performance advantages by building its own chipmaking projects, thus attracting more customers and investors and ultimately increasing its profitability and market attractiveness. However, this also means that Arm might compete directly with its customers by stopping being third-party neutral.

And since chipmaking requires huge capital investment, increasing Arm’s financial burden is not conducive to SoftBank’s initial intention to solve the financial pressure. Arm, who wants to obtain more profits and a higher IPO market valuation in the short term through chipmaking projects, needs to consider more about its chipmaking move.

Enterprise News(April 26

Microsoft and Meta suddenly cut server orders drastically

According to some industry insiders, U.S. cloud service giants Microsoft and Meta have cut more server orders, as Taiwan’s Economic Daily News reported. Microsoft’s annual reduction is up to 30%, with the bulk planned for the second half of the year. The research of a TF International analyst indicated that Intel’s Sapphire Rapids was expected to start mass shipping in 2023, and investors expected this latest server chip to contribute significantly to Intel’s profits. However, the supply chain was recently told to cut Sapphire Rapids chip orders for 2Q23 by about 50% due to weak demand in the server market.
Meta had cut its orders earlier, so its recent cut was less significant than Microsoft’s. However, Meta still cut down on its server orders for 2Q23 by 30-40% and took a conservative view of orders for 2H23. Microsoft and Meta’s drastic order cutting led to unexpected changes in this year’s total server shipments. Initially expected to grow 10% this year, the market could instead enter a recession, putting foundries like Quanta and Wiwynn under pressure.
The original market expectation was that in the wave of large model training driven by ChatGPT, demand for computing power should keep soaring, especially with the release of the fourth generation of GPT. The rapid iteration of large language models has generated a great demand for computing power, and the growth in computing scale has driven the need to enhance single-point computing power for AI training chips. Meeting the demand for computing power involves increasing the number of AI servers or improving the servers’ computing performance. Generally speaking, Microsoft and Meta’s big order cuts might have to do with market demand, technology supply, technology updates, etc.

Microsoft and Meta slashed orders since the demand for the server market remains quite weak. However, in terms of market demand, the AI wave brought by the large language model ChatGPT should be a long-term development direction. The market is unlikely to stop growing, especially now that major countries and global tech giants are launching large model products. From this perspective, maybe Microsoft and Meta cut orders because they had found products with higher server performance. Coincidentally, the industry is accelerating the launch of high-performance AI chips. Hence rapid technology iteration could also be a reason for Microsoft and Meta’s huge order cuts.

Market News(April 26

The DRAM chip market to bottom out and rebound

Samsung Electronics recently announced its plan to cut back memory chip production as increased losses in its memory business dragged down its overall performance. Forced to issue a statement, Samsung said that under the assessment that the company had secured enough volume to respond to future memory demand changes, it would adjust its production lines to lower memory output to a meaningful level and optimize line operations already underway. This is the first time in 25 years since the 1998 financial crisis that Samsung has publicized official production cuts. Original memory makers such as SK Hynix, Micron, Western Digital, and Kioxia started production reductions even earlier. And multiple original manufacturers cut capital expenditure plans for 2023. SK Hynix decreased half from last year, while Micron reduced around 40% from last year.

The industry believes that the first quarter is the year’s biggest off-season, and stocking demand will increase quarter by quarter, which, coupled with the progressive recovery of consumer demand, the gradual relief of inventory pressure, and the further output cuts of original manufacturers, leads to the expectation that this cycle will bottom out in Q2, and the memory market may stabilize in Q3 at the earliest. Original manufacturers’ collective production reductions are starting to show effect. TrendForce estimates that DRAM’s price drop will slow down to 10-15% in Q2, and NAND Flash’s price decrease will shrink to 5%-10%. Analysts said although some DRAM prices may increase in the short term, it was more difficult to see a return to the overall price rise trend given the influences of a high inventory and large module manufacturers having no intention to raise purchase quotations.


As the biggest category within memory chips, DRAM accounts for more than 50% of the global memory market share. The downturn of the consumer electronics market caused a considerable demand reduction for memory products, leading to a downward spiral beyond imagination. Samsung shrank its production in a bid to improve memory oversupply.

In terms of the current market price, DRAM products are already selling at a price lower than the cost price. If taken capacity into account, they are basically at the lowest price in history. However, due to the weak demand in the downstream market, there is no significant increase in sales volume. Major manufacturers such as Samsung, SK Hynix, and Micron can stop inventory increases in the market in the short term by reducing production, thus achieving market expansion and price recovery.

Some Chinese Industrial News

1.Intel CEO visits China, firmly optimistic about the Chinese market

Intel CEO attended Intel’s sustainability summit in Beijing, where he said sustainable development was essential to China’s booming digital economy. Intel looks forward to building an in-depth partnership in sustainability with China, a leader in the global sustainability field.

The five technology superpowers of computing, connectivity, cloud-to-edge infrastructure, artificial intelligence, and sensing and perception are driving the world into a new era of digital economy. In this context, green and sustainable development focusing on carbon and energy reduction has become very important. In the Chinese market alone, Intel conserved 22 million kWh of energy in 2020-2021. Its carbon footprint in China has been reduced by more than 80% in the past 20 years. As reported, Intel joined hands with Lenovo, Dell, HP, ASUS, and other industry ecosystem partners worldwide this year to release the first products based on the green PC concept in China.

Moreover, Intel is also driving sustainable development in China’s three major networks: transportation, power, and 5G. Intel remains bullish on the Chinese market because, in its opinion, the country’s economy continues to develop and prosper nonstop, and its digital economy scale and the consequent demand for computing power are also increasing exponentially.

2.Chip equipment giant’s China sales to surge

As suggested by the comments of California-based Lam Research and Dutch company ASML Holding NV, given the strong demand of China for less advanced products, namely, semiconductor equipment for mature processes, the country may be a bigger customer for the industry than expected this year that focuses on chips used in electric vehicles (EVs). The two companies said that despite the comprehensive restrictions the U.S. had imposed on China’s semiconductor industry last October, these regulations had only affected the equipment used to manufacture the most advanced chips so far. They predict sales to Chinese companies to increase in the coming months.

According to Lam and ASML, Chinese customers are purchasing tools for making less advanced chips for products such as EVs, mobile phones, and PCs. In Lam’s case, it has initially estimated that the restrictions on China would cause it a revenue loss of between US$2 billion and US$2.5 billion in 2023. However, Lam said it had already received a “clarification” of the rules from the U.S. government and an advance of about US$5 billion, primarily from new customers. ASML said it had around €39 billion backlog, equivalent to about two years’ tool shipments. Chinese customers focusing on making less advanced chips account for about 30% of those orders, a jump from last November when ASML said China made up for 18% of its then €38 billion backlog.

3.The U.S. pressures South Korea on chip supply to China

As the U.K.’s Financial Times reported, the U.S. had asked South Korea to urge its chipmakers not to fill any market gap in China if Beijing banned U.S. memory chipmaker Micron from selling chips in China. Yonhap News Agency reported on April 24 that sources from the South Korean Embassy in the U.S. said they were “unaware” of the above information. Some analysts questioned South Korean companies’ capabilities to fill the gap left by Micron as the U.S. government had used provisions in the Chip and Science Act to restrict them from increasing production capacity in China.

The Cyberspace Administration of China (CAC), China’s cyberspace regulator, said in March that it would conduct a cybersecurity review into Micron’s products sold in China. Micron responded that it was collaborating with the Chinese government and that its operations in China were normal. Financial Times quoted sources as saying Washington had asked Seoul to urge Samsung Electronics and SK Hynix not to increase chip sales to China if Micron was banned from selling in China as a result of the investigation. The U.S. has imposed a series of technology export controls on China, fearing that the latter could use those technologies to produce military chips. It has blacklisted several of China’s biggest chipmakers, including Micron’s rival Yangtze Memory Technology.





您的姓名: *
您的公司: *
Email: *
订阅内容: *
  • 公司动态
  • 行业前沿
  • 政策法规
  • 产业洞察