On October 17, 2023, the U.S. government unveiled an updated set of regulations for semiconductor exports, introducing stricter standards for advanced AI chips. Additionally, these regulations expand control over the export of exposure equipment and include Chinese GPU design startups on an Entity List.

TrendForce’s Insights: 

1. New regulations cover chips, manufacturing equipment, and related companies, signaling an effort to restrain China’s AI development.

In this latest set of regulations, the U.S. has relaxed the I/O bandwidth restrictions for AI chips and introduced three additional conditions beyond a total processing performance (TPP) of ≥ 4800 TOPS:

(1) Total processing performance ≥ 1600 TOPS and performance density (PD) ≥ 5.92

(2) Total processing performance ≥ 2400 TOPS but < 4800 TOPS and performance density ≥ 1.6 but < 5.92

(3) Total processing performance ≥ 1600 TOPS and performance density ≥ 3.2 but < 5.92

As a result of these new conditions, NVIDIA’s A800, H800 GPU, and the recent launched L40S GPU for the Chinese market are now included in the list of controlled exports, similar to the A100 and H100 GPUs that were added in September 2022.

Concerning manufacturing equipment, the control threshold for exposure equipment has shifted from single-machine (specified substrate) coverage precision of ≤ 1.5nm to > 1.5nm but ≤ 2.4nm. This change directly led to the inclusion of ASML’s 1980Di DUV lithography machines.

On the corporate front, Chinese domestic GPU design startups such as Birentech, Moore Threads, and high-speed DSP design company Superfusion Semiconductor, along with their related entities, have been placed on the Entity List by the U.S. Department of Commerce.

In summary, these new regulations encompass chips, manufacturing equipment, and related companies. The U.S. is not only controlling the current mainstream AI product lines and applications of DUV lithography machines for 28-7nm processes but is also making a clear effort to interfere Chinese domestic manufacturers’ development of AI computation chips, indicating a strong determination to restrict China’s growth in the AI sector.

2. Priority for Chinese Enterprises: Securing AI Computing Resources through Cloud Service Providers

In light of the impact of the new U.S. semiconductor control regulations, Chinese domestic companies will be limited to AI chip performance not exceeding that of NVIDIA L40 GPU. As leading companies like NVIDIA, AMD, Intel, and others continuously boost the performance of their AI chips, the gap between the AI computing resources established by Chinese companies and their international counterparts will continue to widen.

Looking at it from an angle of independent research and development, with the inclusion of 1980Di and more advanced DUV lithography machines in the control list and the U.S. Department of Commerce placing Chinese IC design companies on the Entity List, short-term mass production of high-performance server AI chips in China seems unlikely.

Faced with challenges in both outsourcing and in-house production, the primary path for Chinese domestic companies to develop AI technology and applications is to obtain high-performance AI computing resources from international cloud service providers (CSP). It is worth noting that the U.S. government is also exploring limitations on Chinese firms attempting to evade semiconductor control policies through CSP. For Chinese companies, establishing robust customer relationships and building extensive AI computing resources are pressing priorities before related policies are enacted.
(Image: Pixabay)

The US Department of Commerce issued new restrictions on AI chips on October 17, 2023, with a focus on controlling the export of chips to China, including NIVIDA’s A800, H800, L40S, and RTX4090, among others. Taiwanese manufacturers primarily serve cloud service providers and brand owners in North America, with relatively fewer shipments to Chinese servers. However, Chinese manufacturers, having already faced two chip restrictions imposed by the US, recognize the significance of AI chips in server applications and are expected to accelerate their in-house chip development processes.

TrendForce’s Insights:

1. Limited Impact on Taiwanese Manufacturers in Shipping AI Servers with H100 GPUs

Major Taiwanese server manufacturering companies, including Foxconn, Quanta, Inventec, GIGABYTE, and Wiwynn, provide AI servers equipped with H100 GPUs to cloud data centers and brand owners in Europe and the United States. These Taiwanese companies have established some AI server factories outside China, in countries such as the US, the Czech Republic, Mexico, Malaysia, and Thailand, focusing on producing L10 server units and L11 cabinets in proximity to end-users. This strategy aligns with the strategic needs of US cloud providers and brand owners for global server product deployment.

On the other hand, including MiTAC, Wistron, and Inventec, also provide server assembly services for Chinese brands such as Inspur and Lenovo. Although MiTAC has a significant share in assembling Inspur’s servers, it acquired Intel DSG (Data Center Solutions Group) business in July 2023. Therefore, the focus of AI servers remains on brand manufacturers using H100 GPUs, including Twitter, Dell, AWS, and European cloud service provider OVH. It is speculated that the production ratio of brand servers will be adjusted before the new restrictions are enforced.

Wistron is a major supplier for NVIDIA’s AI server modules, DGX A100, and HGX H100. Its primary shipments are to end-users in Europe and the United States. It is expected that there will be adjustments in the proportion of shipments to Chinese servers following the implementation of the restrictions.

Compal has fewer AI server orders compared to other Taiwanese manufacturers. It has not yet manifested any noticeable changes in Lenovo server assembly proportions. The full extent of the impact will only become more apparent after the enforcement of the ban.

During the transitional period before the implementation of the chip ban in the United States, the server supply chain can still adapt shipments based on local chip demand in China to address market impacts resulting from subsequent chip controls.

2. Chinese Manufacturers Focusing on Accelerating In-House Chip Development

Chinese cloud companies had already started developing their AI chips before the first U.S. chip restrictions in 2022. This included self-developed AI chips like Alibaba Cloud’s T-HEAD, a data center AI chip, and they expanded investments in areas such as DRAM, AI chips, and semiconductors with the aim of establishing a comprehensive IoT system from chips to the cloud.

Baidu Cloud, on the other hand, accelerated the development of its third-generation self-developed Kunlun chip, designed for cloud and edge computing, with plans for an early 2024 release.

Tencent introduced three self-developed chips in 2021, including an AI inference chip called Zixiao, used for Tencent’s meeting business; a video transcoding chip called Canghai, used in cloud gaming and live streaming applications; and a smart network card chip named Xuanling, applied in network storage and computing.

ByteDance made investments in cloud AI chips through its MooreThread initiative in 2022 for applications in AI servers. Huawei released the Ascend 900 chip in 2019 and is expected to introduce the Ascend 930B AI chip in the latter half of 2024. While this chip has the same computational power as the NVIDIA A100 chip, its performance still requires product validation, and it is speculated that it may not replace the current use of NVIDIA GPUs in Chinese AI servers.

Despite the acceleration of self-developed chip development among Chinese cloud server manufacturers, the high technological threshold, lengthy development cycles, and high costs associated with GPU development often delay the introduction of new server products. Therefore, Chinese cloud companies and brand manufacturers continue to purchase NVIDIA GPUs for the production of mid to high-end servers to align with their economic scale and production efficiency.

In response to the new U.S. restrictions, Chinese cloud companies have adopted short-term measures such as increasing imports of existing NVIDIA chips and building up stockpiles before the enforcement of the new restrictions. They are also focusing on medium to long-term strategies, including accelerating resource integration and shortening development timelines to expedite GPU chip manufacturing processes, thus reducing dependency on U.S. restrictions.

According to a report by Bloomberg, Yoshihiro Seki, Secretary-General of the ruling Liberal Democratic Party and a member of the Japanese parliament, has announced that the government is planning to allocate an additional ¥900 billion for the construction of TSMC’s Fab 2 in Kumamoto, Japan. Furthermore, an extra ¥590 billion in subsidies will be provided to support the construction of a wafer fab by the Japanese semiconductor startup Rapidus.

Seki emphasized that subsidies usually cover about one-third of the total investment. With measures like training Japanese engineers and collaborative R&D with local companies, this subsidy could increase to potentially cover up to half of the investment. He also noted that the specific amount remains subject to change as the additional budget has not been finalized yet.

The Japanese government initiated the “Strategy for Semiconductors and the Digital Industry” in 2021 to address economic risks and prepare for the wave of digitalization. At that time, they already provided ¥476 billion in subsidies for TSMC’s Kumamoto 1st Fab. The current subsidy marks an expansion of these efforts.

The local government Kumamoto is eagerly anticipating TSMC’s presence. Ikuo Kabashima, the Governor of Kumamoto Prefecture, recently proposed “New Airport Concept Next Stage” that envisions using the airport as a hub for semiconductor imports and exports over the next decade. This plan aims to stimulate the clustering of semiconductor-related industries and contribute to regional development centered around Kumamoto.

Moreover, the Japanese government has pledged to provide ¥330 billion in funding to enable Rapidus to construct a 2nm wafer fab in Hokkaido. These substantial subsidies underscore the Japanese government’s commitment to these semiconductor projects.

In response to the Japanese government’s additional subsidies, Tetsuro Higashi, Chairman of Rapidus, stated in an interview with Jiji Press on the 24th that apart from the new factory being built in Chitose, Hokkaido, “We also plan to construct second and third factories, and they will also be situated in Chitose, Hokkaido.” Rapidus’s 2nm chip R&D/production facility, Chitose Fab IIM-1, located in the Chitose Meimeimei World industrial park in Chitose, Hokkaido, commenced construction in September. The trial production line is expected to start in April 2025, with mass production slated to begin in 2027.

The semiconductor foundry, United Microelectronics Corporation (UMC), held an online briefing on October 25th to unveil its 3Q 2023 operational report. UMC achieved consolidated revenue of NT$57.07 billion, marking a 1.4% growth compared to the previous quarter’s NT$56.3 billion in 3Q23. However, it’s essential to note that this quarter’s revenue decreased by 24.3% in comparison to 3Q 2022.

In 3Q, a 35.9% gross margin yielded a net profit of NT$15.97 billion and an EPS of NT$1.29. The first three quarters of 2023 saw revenue at NT$167.575 billion, marking a 20.5% decline from 2022. The gross margin for this period remained at 35.8%, resulting in a net profit of NT$47.795 billion and an EPS of NT$3.87.

UMC’s Co-president, Jason Wang, highlighted that the company’s performance in the 3Q was boosted by the growing demand in the computer and communication sectors. This was further enhanced by ongoing improvements in product offerings and favorable exchange rates. Notably, despite a 2.3% decrease in overall wafer shipments, the revenue and gross margin remained robust compared to the previous quarter.

Delving into the terminal product market, products like LCD controllers, Wi-Fi, encoders and decoders, and touch IC controllers stimulated demand in the computer application sector. Additionally, the demand for RF front-end ICs and network chips contributed to the shipment volume in the communication sector.

Looking ahead to the 4Q, Wang said that the computer and communication sectors are gradually recovering in terms of short-term demand. In contrast, the automotive market remains challenging, and customers are adopting a cautious approach in managing inventory levels.

UMC foresees that the expansion of capacity at Fab 12A P6 in Nanjing in 2024 will provide significant support, further boosting revenue contributions for 22/28-nanometer technologies.

UMC’s estimate for the 4Q indicates that wafer shipments are projected to decline by 5%, with the average selling price remaining stable. Capacity utilization is expected to decrease from 67% in the previous quarter to a range of 61-63%, which will consequently impact the gross margin. It is estimated to decrease from 35.9% in the 3Q to a range of 31-33%.

Regarding capital expenditure, Q3 saw approximately $570 million spent, a 30.49% decrease from the previous quarter and a 25.39% decrease from 3Q 2022. Cumulative capital expenditure for the first three quarters reached around $2.4 billion, showing a 52.69% increase compared to 2022. The total 2023 capital expenditure remains at $3 billion, with 90% allocated to 12-inch capacity and 10% to 8-inch capacity.

(Image: UMC)

SK hynix today reported the financial results for the third quarter ended September 30, 2023. The company recorded revenues of 9.066 trillion won, operating losses of 1.792 trillion won and net losses of 2.185 trillion won in the three-month period. The operating and net margins were a negative 20% and 24%, respectively.

After bottoming out in the first quarter, the business has been on a steady recovery track, helped by growing demand for products such as high-performance memory chips, the company said.

“Revenues grew 24%, while operating losses narrowed 38%, compared with the previous quarter, thanks to strong demand for high-performance mobile flagship products and HBM3, a key product for AI applications, and high-capacity DDR5,” the company said, adding that a turnaround of the DRAM business following two quarters of losses is particularly hopeful.

SK hynix attributed the growth in sales to increased shipments of both DRAM and NAND and a rise in the average selling price.

By products, shipments of DRAM increased 20% from the three months earlier, thanks to strong sales of high-performance products for server applications such as the AI with the average selling price also recording a 10% rise. Shipments of NAND also rose with high-capacity mobile products and solid state drive products taking the lead.

Following a turnaround, an improvement in the DRAM business is forecast to gain speed, backed by popularity of the generative AI technology, while there are looming signs of a steady recovery in the NAND space as well.

With the effect of the production reduction by global memory providers starting to be seen and customers, following efforts to reduce inventories, placing new orders now, semiconductor prices are starting to stabilize, the company said.

To meet new demands, SK hynix plans to increase investments in high-value flagship products such as HBM, DDR5, and LPDDR5. The company will increase the share of the products manufactured from the 1anm and 1bnm, the fourth and the fifth generations of the 10nm process, respectively, while increasing investments in HBM and TSV.

(Image: SK hynix)