Despite tightening U.S. export controls on AI chips, China seems to be making strides in its AI development, fueled by strong government investment and local expansion. A report from Economic Daily News, citing Financial Times, suggests Huawei is aiming to expand its market share by promoting self-developed chips for AI inference tasks, encroaching on NVIDIA’s territory.

On the other hand, China’s government-backed National Integrated Circuit Industry Investment Fund (Big Fund) is advancing its third phase with a ¥344 billion ($47 billion) initiative to strengthen its semiconductor industry. Key focus areas may include AI computing chips and HBM chips, according to Commercial Times.

Huawei Tries to Turn the Tide by Expanding AI Inference Business

Major AI companies in China used to rely heavily on NVIDIA’s GPUs for training large language models. However, as model training gradually slows and AI applications like chatbots gain wider adoption, demand for inference will likely increase, creating opportunities for Huawei.

The Financial Times report notes that the Chinese tech giant now focuses on a more technically feasible path. Though incapable of competing with NVIDIA in the AI training sector, it is reportedly positioning its latest Ascend AI chips as the hardware of choice for “inference” tasks.

By running NVIDIA-trained AI models on its Ascend chips. Huawei is helping local businesses bridge the gap with compatibility software tools, the report adds.

Though NVIDIA’s CUDA is user-friendly and accelerates data processing, Huawei’s upcoming Ascend 910C chip is expected to address concerns, as the new hardware will be paired with improved software to make it easier for developers to use, as per Financial Times.

China’s USD 47 Billion Big Fund Gears up

On the other hand, amid U.S. efforts to limit China’s AI development, China’s government-backed “National Integrated Circuit Industry Investment Fund” (Big Fund) has been actively pushing forward its third phase.

According to Commercial Times, aside from previous focuses on semiconductor manufacturing, equipment, materials, and components, the rapid rise of AI technology in the past two years may shift investment attention to key AI semiconductor areas, such as computing chips and HBMs.

The third phase of the Big Fund, registered on May 24, 2024, has a capital of ¥344 billion (USD 47 billion), higher than the combined total of the first and second phases, the report adds.

As reported by TrendForce, this ambitious program focuses on supporting domestic chip production equipment developers, particularly as access to advanced tools from companies like ASML becomes increasingly restricted. Managed by Huaxin Investment Management, the fund’s initial investment of ¥93 billion ($12.685 billion) targets key material and equipment manufacturers.

However, TrendForce notes that this investment may fall short of enabling China to compete with established market leaders. Since its inception in 2014, the Big Fund’s efforts have faced significant headwinds, particularly due to U.S. sanctions that have disrupted the operations of major Chinese semiconductor firms.

Read more

• [News] Huawei Reportedly Begins Sampling of Ascend 910C to Rival NVIDIA
• [News] Huawei’s Self-developed AI Chip Challenges NVIDIA, Boasting Its Ascend 910B to Be Equal in Match With A100

(Photo credit: Huawei)

Ahead of SK hynix’s upcoming fourth-quarter results announcement, another Korean memory giant, Samsung, may be in trouble again. According to MoneyToday, the company has delayed the development of its 10nm-class sixth-generation (1c) DRAM by another six months, postponing the target date to June, 2025.

SK hynix May Start 1c DRAM Mass Production in February

On the other hand, MoneyToday revealed in a previous report that SK hynix is set to become the first in the world to mass-produce DRAM using the 1c process as soon as in February.

According to the report, SK hynix has recently completed mass production qualification for 1c DDR5.

Samsung’s HBM4 Production Faces Potential Delay

However, Sources told MoneyToday that while Samsung secured its first fully functional 1c DRAM chip in late 2024, it fell short of the target yield, typically 60-70% for mass production readiness.

Notably, the delay, caused by yield issues, may also affect Samsung’s planned HBM4 mass production in the second half of 2025, the report adds.

According to the report, the industry typically considers an 18-month development cycle for each generation. While Samsung developed its 5th-generation 10nm-class (1b) DRAM in December 2022 and announced mass production in May 2023, there has been no news regarding its 1c DRAM progress since then.

The hope of catching up may becoming even more distant for Samsung, as its HBM archrival has seem to be progressing on track. According to a previous report from ZDNet, SK hynix is planning to ship HBM4 samples to NVIDIA as early as June 2025, with full-scale product supply expected to begin around the end of the third quarter.

The MoneyToday report brings up concerns that the delay in 1c DRAM development will certainly affect Samsung’s HBM plans. While DDR5 is the core product for 1c, HBM development typically follows later. If 1c DRAM production begins at the end of 2025, HBM4 production may be further pushed into 2026, as highlighted by MoneyToday.

To avoid the worst-case scenario, the report indicates that Samsung is making adjustments to the 1c DRAM design, and will do its best to speed up the development process.

Read more

• [News] Samsung Has Reportedly Ordered Equipment for 1c DRAM Mass Production, Paving the Way for HBM4

(Photo credit: Samsung)

According to a report from MoneyDJ, citing Sammobile, semiconductor intellectual property (IP) leader Arm is reportedly planning a significant price increase of up to 300%. This move could severely impact the future development of Samsung Electronics’ Exynos series processors, as highlighted in the report.

As Sammobile notes, Samsung is currently facing several challenges. One major issue is the low yield rate of its 3nm manufacturing process, which has hindered its foundry division’s ability to secure external orders. Additionally, due to this yield issue, the upcoming Galaxy S25 flagship smartphone will not feature Samsung’s in-house Exynos 2500 chipset. Instead, it will rely on Qualcomm’s Snapdragon 8 Elite processor.

Arm’s Reported Price Hike: Major Impact on Samsung’s Exynos Chips

Citing Reuters, the MoneyDJ report highlights market rumors that Arm intends to raise prices by as much as 300%. According to Sammobile, this poses a significant challenge for Samsung as it primarily purchases ready-made CPU cores from Arm. The Exynos 2400 chipset features a 10-core CPU based on Arm’s Cortex-X4, A720, and A520 cores.

If Arm raises its prices, Samsung’s costs for purchasing CPU cores are expected to skyrocket, significantly affecting its flagship Exynos chips used in premium smartphones.

The report highlights that Arm’s reported price increase could create substantial difficulties for Samsung, particularly as most Android OEMs favor Qualcomm’s Snapdragon processors or MediaTek’s Dimensity processors over Samsung’s Exynos chips.

Samsung’s Lack of Custom CPU Cores Poses Disadvantages

By contrast, Qualcomm’s Snapdragon processors no longer use Arm-designed cores, as highlighted in the report. The Snapdragon 8 Elite features Qualcomm’s custom-designed Oryon cores. While Qualcomm does utilize Arm’s licensed instruction set architecture (ISA) to maintain compatibility with Arm-based software such as Android, the report points out that Arm’s reported price hike will have a far smaller impact on Qualcomm compared to Samsung.

As noted in the report, Samsung previously developed custom cores for its Exynos chipsets, such as the M5 custom core used in the Exynos 990, which powered the Galaxy S20 and Galaxy Note 20 in 2020.

However, consistent underperformance of Exynos chips compared to Snapdragon processors led Samsung to dissolve its CPU development team and shift to sourcing ready-made cores directly from Arm, as the report states.

Uncertain Future for Samsung’s Exynos Chips

The Sammobile report indicates that it is unlikely Samsung will resume developing custom cores, as this would require years of effort and substantial investment. In fact, the report points out that it might actually be a more cost-effective solution for Samsung’s Mobile eXperience (MX) division to directly purchase Snapdragon processors from Qualcomm.

Read more

• [News] Samsung Faces Profit Challenges as Huawei Joins Apple in Achieving Tech Independence
• [News] Samsung Electronics Anticipates Mounting Challenges Amid Faltering Chip Business

(Photo credit: Samsung)

According to a report from MoneyDJ, citing Japanese media outlet Yomiuri Shimbun, Japanese auto giant Nissan Motor plans to implement layoffs in Japan. The company is set to halt production of a commercial vehicle model at its Shonan plant in Kanagawa Prefecture, leading to hundreds of job cuts.

The report highlights that Nissan had previously announced a global workforce reduction plan of 9,000 employees in November 2024, and this marks the first time the company has disclosed specific layoff measures within Japan.

Nissan has reportedly informed its suppliers that the Shonan plant will cease production of the “AD” commercial vehicle in November this year. The decision comes amid sluggish sales, attributed to increasing competitive pressure from rivals like Toyota, as indicated by the report.

The Shonan plant, which has an annual production capacity of 150,000 units, currently manufactures the AD and NV200 models. The report states that Nissan is also reportedly considering scaling down production of the NV200.

Nissan’s Financial Woes and Their Impact on Business Integration Negotiations with Honda

Nissan’s ongoing financial struggles are seen as a factor that could influence its business integration talks with Honda. According to the report, Nissan and Honda announced on December 23, 2024, that they had signed a basic agreement to begin discussions and evaluations regarding a potential merger.

The two companies aim to reach a final merger agreement by June 2025 and plan to establish a joint holding company by August 2026. This holding company would oversee both Nissan and Honda as wholly-owned subsidiaries and is scheduled to list on the Tokyo Stock Exchange Prime Market in August 2026. At that time, both Nissan and Honda would be delisted from the exchange.

However, the negotiations face several challenges. The report, citing Kyodo News, notes that Honda has requested Nissan to repurchase shares held by Renault, the French automaker, to prevent those shares from being acquired by third parties, which could potentially disrupt the merger discussions.

The report further highlights that Taiwan’s Foxconn has reportedly approached Renault with an interest in acquiring its Nissan shares. Foxconn’s potential involvement has become another key factor in the ongoing merger negotiations.

Citing Jiji Press, the report states that Honda and Nissan’s planned joint holding company is expected to receive the majority of its capital contribution from Honda, given its significantly higher market capitalization compared to Nissan.

Meanwhile, Nissan’s financial difficulties pose a significant risk to the integration, as highlighted in the report. The complex interplay of factors—including Nissan’s weak business performance, Renault’s stake in the company, and potential third-party involvement such as Foxconn—further adds layers of uncertainty to the merger process.

Read more

• [News] Honda Reportedly Urges Nissan to Acquire Renault Stake to Prevent Third-Party Influence in Merger
• [News] Honda-Nissan Merger Talks Reportedly Spurred by Concerns over Foxconn’s Potential Acquisition

(Photo credit: Nissan)

According to a report from Commercial Times, sources indicate that due to U.S. restrictions on AI computing power exports to China, the application for an ADAS (Advanced Driver Assistance System) chip by Taiwanese ASIC manufacturer Alchip was rejected.

The ADAS chip developed by Li Auto, named “Schumacher,” was co-developed with Alchip and originally planned for production using TSMC’s 5nm process. The chip was scheduled for tape-out at the end of last year, but the application was denied, and the project is now undergoing redesign, as stated in the report.

The report highlights that automotive chip manufacturers believe that, under the current trends for end-to-end large models, automotive intelligent driving chips with approximately 800 TOPs of computing power are considered industry-leading. It is estimated that excessively high computing power is unnecessary, with the focus instead placed on computing power requirements for the back-end training.

The report further points out that companies such as NVIDIA and Qualcomm offer autonomous driving solutions. However, to avoid potential supply chain issues, such as shortages or sanctions, major companies centered around intelligent driving, including NIO and XPeng Motors, are actively developing their own chips.

Alchip’s Focus on Co-Packaged Optics (CPO)

As noted by the report, supplier selection for AWS Trainium 4 has commenced. AWS is reportedly considering the use of SoIC (System on Integrated Chips) and CPO (Co-Packaged Optics) architectures for its Trainium 4 ASIC.

Notably, Alchip is developing related chip designs for AWS and has successfully completed the design finalization of a 2nm SoIC chip, which could position the company to secure next-generation chip orders from major clients, as the report indicates.

The report underscores that, from the perspective of major CSP (Cloud Service Provider) companies, CPO is expected to become a central focus for AI ASICs used in training and inference tasks. Similarly, general-purpose GPU manufacturers are increasingly focusing on silicon photonics, aligning with this emerging trend.

Read more

• [News] U.S. Adds Tech Giants CXMT, Tencent, and CATL to List of Chinese Firms Linked to Military Support
• [News] A Quick Recap: Impact of the Latest U.S. Chip Export Curbs on China and Global Semiconductor Giants

(Photo credit: Li Auto)

A 6.4 magnitude earthquake struck southern Taiwan early Tuesday, at 12:17 a.m., raising concerns about its impact on foundry giant TSMC. According to Commercial Times, the Economic Daily News and Reuters, TSMC evacuated staff from its central and southern Taiwan facilities, ensuring all employees were safe.

The reports, citing TSMC, notes that the extent of the impact is currently under assessment.

Notably, TSMC’s fab 18 in the Southern Taiwan Science Park is a key production site for its advanced nodes, featuring 3nm.

The quake, which was also felt in Taipei, caused buildings to shake shortly after midnight. With a depth of 9.4 km, it was centered in Dapu Township, Chaiyi, Southern Taiwan. Nearby fabs include TSMC’s Fab8, Fab14, Fab18, and UMC’s Fab12A in Tainan, as well as TSMC Fab15 in Taichung.

According to TSMC, the largest tremor recorded at its fabs in the Southern Taiwan Science Park was level 5, while the Central Taiwan Science Park recorded a level 4 tremor. The fabs in the Hsinchu Science Park, including those in Longtan and Zhunan, experienced a level 3 tremor.

As its fabs in the Central and Southern Taiwan Science Parks met evacuation thresholds, the company implemented emergency response procedures, evacuating personnel indoors and outdoors, and is conducting safety checks, as highlighted by a Technews report.

According to the Economic Daily, since most of TSMC’s equipment is designed with anti-vibration features, significant damage is not expected. However, the prolonged shaking raises concerns about potential equipment misalignment, requiring shutdowns for repositioning and recalibration before restarting, the report adds.

On the other hand, industry insiders cited by Commercial Times revealed that following the earthquake, supply chain partners and engineers across Taiwan have been mobilized to TSMC’s facilities in Central and Southern Taiwan Science Parks to conduct recovery efforts, aiming to restore production as soon as possible.

Read more

• [News] TSMC Reportedly Building Two More CoWoS Facilities, Debunking Order Cut Rumors

(Photo credit: TSMC)

According to a report from The Elec, Apple is planning to launch a MacBook Air with an oxide thin-film transistor (TFT) liquid crystal display (LCD) in 2027, while the plan of implementing OLED screens on the MacBook Air has been delayed.

As highlighted in the report, this will be the first time Apple incorporates an oxide TFT LCD into its MacBook Air lineup. The report also notes that LG Display is likely to be the supplier for these oxide TFT LCDs, since LG Display and Sharp have been major suppliers of LCD for MacBooks.

MacBook products are divided into two tiers: the higher-tier Pro models and the lower-tier Air models. Apple’s 14.2-inch and 16.2-inch MacBook Pros, released in 2022, feature oxide TFT LCD displays. However, the 13.6-inch and 15.4-inch MacBook Air models released in the same year used amorphous silicon (a-Si) TFT LCDs, which, as noted in the report, have slower electron mobility compared to oxide TFT technology.

Regarding the delay of the OLED MacBook Air, the report indicates that it may be linked to the underwhelming sales performance of OLED iPads released in 2024.

OLED MacBook Airs could potentially launch around 2029, as indicated by a report from GSMArena.

Read more

• [News] Apple Reportedly Nears Final Check on TSMC’s U.S.-made Chips, with 4nm Mass Production Imminent
• [News] Four Highly Anticipated Apple Products Set to Launch in Early 2025

(Photo credit: Apple)

As the Biden administration brought hammer down on AI chip exports by introducing a three-tier system based on countries and companies, NVIDIA has voiced its disagreements on the regulations, calling it a “misguided AI diffusion rule.

Notably, in a recent 8K filing with the SEC, the U.S. chipmaker outlined its assessment of the impact of these new restrictions, highlighting that its latest B200 and GB200 products will be affected.

In a 8K filing it submitted to SEC on January 13rd, NVIDIA revealed the scope of products affected by the AI diffusion rule. According to its statement, the rule will impact a broad range of its GPU products starting May 15, 2025, unless changes are made.

The rule applies to products meeting 3A090.a or 4A090.a specifications, including A100, A800, H100, H200, H800, B100, B200, GB200, L4, L40S, and RTX 6000 Ada, as noted by NVIDIA.

No Material Financial Impact Mentioned

It is worth noting that the filing does not warn of any material financial impact. While the new rules won’t take effect for 120 days, it may allow the incoming Trump administration time to weigh in.

Unlike previous rules, which are mostly presented as blacklists, the new restrictions apply globally, using a whitelist approach. Only 18 key allies and partners are exempt from the controls, with countries like U.S., Canada, Germany, Japan, the Netherlands, South Korea and Taiwan leading the way.

Furthermore, in terms of the “Additional Due Diligence Measures” Rule, which requires additional diligence for foundries and OSATs to verify technical parameters, NVIDIA states that as it is an approved IC designer, the rule is not expected to have an impact on its operations or on product sales to China.

Uncertainties Lurk as China Remains a Major Market

However, according to a previous report from Reuters, NVIDIA earns about 56% of its revenue from international markets, with China accounting for 17% of sales. Therefore, the new rules will likely reduce NVIDIA’s market in a considerable way, as nearly half of its chips now go to countries that will be restricted, as noted by the report.

Additionally, the Reuters report notes that the new rules may allow major cloud providers like Microsoft, Google, and Amazon to seek approval to bypass licensing for AI chips, enabling them to set up data centers in restricted countries, which will boost their market dominance in AI.

Read more

• [News] Biden’s AI Chip Export Curbs in the Works: Key Points of Three-Tier System at a Glance
• [News] NVIDIA’s Revenue at Risk as U.S. AI Chip Export Restrictions Loom; Big CSPs Set to Benefit

(Photo credit: NVIDIA)

According to a report from Economic Daily News, TSMC plans to build two new CoWoS facilities at the Phase III site of Southern Taiwan Science Park (STSP), with an estimated investment exceeding NTD 200 billion.

Construction at the STSP Phase III site is expected to begin as early as March 2025, and the two new facilities are projected to be completed by April 2026, with equipment installation likely commencing shortly thereafter, as the report notes.

As indicated by the report, the Southern Taiwan Science Park Administration has confirmed TSMC’s submission of a land lease application but declined to disclose details about the company’s specific development plans. Meanwhile, TSMC Chairman C.C. Wei reaffirmed the company’s commitment to expanding CoWoS production capacity during last week’s earnings call.

Citing industry sources, the report suggests that the new CoWoS facilities at the STSP will occupy a land area of 25 hectares, surpassing the 20-hectare of the ChiaYi facility.

NVIDIA’s Order of CoWoS-L Capacity

Recently, speculation about NVIDIA cutting its CoWoS orders from TSMC has been circulating, as the report points out. However, NVIDIA CEO Jensen Huang dismissed these rumors during his recent visit to Taiwan. Huang stated there has been no reduction in CoWoS demand and revealed an increasing need for CoWoS production, particularly for CoWoS-L capacity.

According to TrendForce, looking ahead to 2025, NVIDIA plans to strategically promote the B300 and GB300 lines—which utilize CoWoS-L technology—thereby boosting the demand for advanced packaging solutions.

TrendForce estimates that, driven primarily by demand for NVIDIA’s Blackwell, the share of CoWoS-L is expected to increase from around 15% in 2024 to 50-60% in 2025, becoming the dominant segment.

TSMC’s CoWoS Expansion Plan Update

TSMC plans to expand eight CoWoS facilities in the short term, as the report notes. This includes two at ChiaYi Science Park Phase 1 and two facilities acquired from Innolux (AP8). Notably, although ChiaYi Science Park Phase 2 was initially slated to host two additional facilities, the land there will not be available until January 2026 at the earliest, as indicated by the report. As a result, TSMC has reportedly prioritizes constructing two CoWoS facilities at the STSP Phase III site. The locations for the remaining two planned facilities are still under evaluation.

Read more

• [News] TSMC Faces Order Cut Fears as AMD, Broadcom, and NVIDIA Reportedly Slash CoWoS-S Demand
• [News] TSMC’s Advanced Process and CoWoS Expansion: Progress Unveiled
• NVIDIA Renames Blackwell Ultra to B300 Series; CoWoS-L Expected to See Growth by 2025, Says TrendForce

(Photo credit: TSMC)

Ahead of Trump’s inauguration, U.S. semiconductor giant Intel is rumored to be connected with more potential buyers for its struggling chip business. A report from SemiAccurate discloses that a company is interested in buying Intel outright, not just parts, and this isn’t one of the firms mentioned in rumors after former CEO Pat Gelsinger’s departure.

The report even claims that SemiAccurate has received direct confirmation from another high-level source, increasing its confidence in the plan from about 60% to over 90%.

So, who would be the key figure to “Make America Great Again” by acquiring Intel? Here is a quick roundup of recent speculations and related analysis compiled by TrendForce.

Elon’s Musk’s Next Step?

Interesting enough, there were rumors indicating that at the weekend, private jets from Musk, executives of Qualcomm, and GlobalFoundries have gathered at Trump’s Mar-a-Lago Club.

The man who eagerly explores in the space with his rockets now might be showing interest in microchips. Another Tipranks report, citing Citi, claims that Elon Musk was mentioned as a potential buyer of Intel on SemiAccurate. However, the report also brings up concerns that this may appear to be a “poor outcome” for Intel unless Team Blue hires a new CEO with expertise in advanced manufacturing, technical skills, and Intel experience.

GlobalFoundries to Team up?

Notably, another report from Bloomberg suggests that senior government officials are considering a potential deal between Intel and GlobalFoundries, while some see GlobalFoundries as a strong partner due to its status as a trusted Pentagon supplier. In September, 2024, Intel secured a $3.5 billion Pentagon contract to produce advanced semiconductors for military and intelligence use.

As they are both major U.S.-based foundries, a merger between GlobalFoundries and Intel could create a company making everything from basic semiconductors to advanced processors, the report adds.

However, the merger could pose challenges. For one thing, Mubadala Investment Co., the Abu Dhabi government’s investment arm, owns about 80% of GlobalFoundries, as noted by Bloomberg. For another, GlobalFoundries moved away from leading-edge node production years ago due to a lack of orders, and it may not have the funds for an acquisition, the report indicates.

Qualcomm and Broadcom May be Moving Away

The ones with the capital, however, may no longer be interested in the deal. A previous Bloomberg report noted that Qualcomm’s interest in acquiring Intel has waned, as the complexity of acquiring all of Intel has made the deal less appealing. However, sources suggested Qualcomm could reportedly consider buying parts of Intel or revisit the idea later.

On the other hand, Broadcom, known for its custom ASICs co-developed with big CSPs, told the Financial Times that the company is fully focused on advancing its AI business and has no intention of acquiring Intel.

Read more

• [News] Intel Reportedly Confirms Panther Lake to Be Built on 18A, with Samples Distributed to Customers
• [News] Broadcom Reportedly Focused on AI Development, Has No Plans to Acquire Intel

(Photo credit: Intel)