Intel, Qualcomm, Google, and other tech giants are reportedly joining forces with over a hundred startups to challenge NVIDIA’s dominance in the market, as per a report from Reuters. Reportedly, their goal is to collectively penetrate the artificial intelligence (AI) software domain, guiding developers to migrate away from NVIDIA’s CUDA software platform.

NVIDIA’s CUDA is a parallel computing platform and programming model designed specifically to accelerate GPU computing. It allows GPU users to fully leverage their chip’s computational power in AI and other applications. As per a previous report from TrendForce, since 2006, NVIDIA has introduced the CUDA architecture, nearly ubiquitous in educational institutions. Thus, almost all AI engineers encounter CUDA during their academic tenure.

However, tech giants are now reportedly aiming to disrupt the current status quo. According to a report from Reuters on March 25th, Intel, Qualcomm, and Google are teaming up to challenge NVIDIA’s dominant position. They plan to provide alternative solutions for developers to reduce dependence on NVIDIA, encourage application migration to other platforms, and thereby break NVIDIA’s software monopoly and weaken its market influence.

The same report from Reuters further indicated that several tech companies have formed the “UXL Foundation,” named after the concept of “Unified Acceleration” (UXL), which aims to harness the power of acceleration computing using any hardware.

The project plans to leverage Intel’s oneAPI technology to develop software and tools supporting multiple AI accelerator chips. The goal is to reduce the technical barriers developers face when dealing with different hardware platforms, streamline the development process, enhance efficiency, and accelerate innovation and application of AI technology.

Vinesh Sukumar, Head of AI and Machine Learning Platform at Qualcomm, stated, “We’re actually showing developers how you migrate out from an NVIDIA platform.”

Bill Magro, Head of High-Performance Computing at Google, expressed, “It’s about specifically – in the context of machine learning frameworks – how do we create an open ecosystem, and promote productivity and choice in hardware.” The foundation is said to aim to finalize technical specifications in the first half of this year and strives to refine technical details by the end of the year.

However, CUDA software has established a solid foundation in the AI field, making it unlikely to be shaken overnight. Jay Goldberg, CEO of financial and strategic advisory firm D2D Advisory, believes that CUDA’s importance lies not only in its software capabilities but also in its 15-year history of usage. A vast amount of code has been built around it, deeply ingraining CUDA in numerous AI and high-performance computing projects. Changing this status quo would require overcoming significant inertia and dependency.

Read more

• [News] NVIDIA’s EULA Amendments Tighten Grip, Suppressing Third-Party CUDA Emulation
• [Insights] Microsoft Unveils In-House AI Chip, Poised for Competitive Edge with a Powerful Ecosystem

(Photo credit: NVIDIA)

Apple is reportedly intensifying its AI efforts, with plans to collaborate with Baidu on the Chinese version of its iPhone 16 series this year, as per a report from Economic Daily News. The devices will feature Baidu’s developed Generative AI technology, thus sparking a new wave of AI server deployment by Baidu.

In China, due to official requirements, providing generative AI services to ordinary consumers requires prior approval and scrutiny from relevant authorities. Only Chinese companies like Baidu and Alibaba have been granted permission. Apple has chosen Baidu as its AI service partner in the Chinese market.

According to the aforementioned reports, Apple and Baidu have entered into negotiations, planning to integrate Baidu’s generative AI services into products such as the iPhone for sale in China. By incorporating a Chinese version of AI, Apple aims to enhance its competitive advantage in the Chinese market.

Inventec, a Taiwanese manufacturer, has had close collaboration with Baidu for 13 years in the field of customized server manufacturing, has jointly developed an AI computing platform. Serving as a key server manufacturing partner for Baidu’s “All in AI” strategy, Inventec plays a major role in Apple’s swift push into the Chinese AI market alongside Baidu.

Inventec has traditionally refrained from commenting on order dynamics and customer relationships, emphasizing instead the robust shipment momentum of AI servers this year. Shipments are expected to more than double compared to last year. With AI server revenue accounting for approximately 5% to 6% last year, it is projected to surpass 10% this year.

Sources cited by the same report from Economic Daily News are optimistic that with Baidu becoming a key partner for Apple in the China’s generative AI landscape, Inventec’s AI server shipment momentum is poised to amplify, benefiting from the opportunities brought by the collaboration between Apple and Baidu.

As per a previous report from TrendForce, Baidu’s foray into AI chips can be traced back to as early as 2011. After seven years of development, Baidu officially unveiled its self-developed AI chip, Kunlun 1, in 2018. Built on a 14nm process and utilizing the self-developed XPU architecture, Kunlun 1 entered mass production in 2020. It is primarily employed in Baidu’s search engine and Xiaodu businesses.

Read more

• [News] Apple Reportedly Places Large Orders for TSMC’s 3nm While Simultaneously Securing Significant Advanced Packaging Capacity
• [News] Boosting Sales? Rare Price Cuts on Apple China’s Official Website

(Photo credit: Apple)

Benefited from frequent orders from its top three clients, Apple, Intel, and AMD, strong momentum has reportedly been driven for TSMC’s 3-nanometer orders, as per a report from Economic Daily News. Anticipated to see sequential growth throughout the year, these orders are expected to remain robust until the end of the year, positioning TSMC as a leader in the semiconductor industry’s recovery.

TSMC has a longstanding policy of not commenting on customer order dynamics. However, it is cited in the report that in the fourth quarter of last year, 3-nanometer orders accounted for approximately 15% of its revenue. With the adoption of 3-nanometer production by major clients this year, revenue from 3-nanometer orders is expected to surpass 20%, becoming the second-largest revenue contributor, following only the 5-nanometer process.

Looking at the orders placed by the top three clients for the 3-nanometer process, Apple is set to introduce the A18 series processor in its iPhone 16 lineup this year. Additionally, the latest self-developed M4 chip for laptops will also be produced by TSMC using the 3-nanometer process, starting in the second quarter.

On the Intel side, the Lunar Lake central processor, graphics processor, and high-speed IO chip are all confirmed to begin mass production at TSMC in the second quarter. This marks Intel’s first instance of outsourcing its entire mainstream consumer platform chip series to TSMC, making it a significant new source of orders for TSMC’s 3-nanometer process this year.

AMD, on the other hand, is poised to unveil its new Zen 5 architecture platform under the code name “Nirvana” this year, expected to significantly enhance AI applications. Following its customary practice, AMD will utilize TSMC’s wafer foundry services, with production set to commence on the 3-nanometer process and an expected launch in the latter half of the year.

TSMC, reportedly, is expanding its production capacity for the 3nm family and advanced packaging this year to meet the large orders from major clients such as Apple, NVIDIA, and AMD in the coming years.

As per TrendForce’s data, the 3nm process alone contributed 6% to TSMC’s Q3 revenue, with advanced processes (≤7nm) accounting for nearly 60% of its total revenue.

Read more

• [News] Apple Reportedly Places Large Orders for TSMC’s 3nm While Simultaneously Securing Significant Advanced Packaging Capacity
• [News] TSMC Reportedly Attracts Influx of 3nm Customers, Predicts Production Capacity Utilization to Surge to 80% by Year-End

(Photo credit: TSMC)

Weak demand in mature process has triggered another wave of foundries’ price reductions. According to a report from Economic Daily News citing industry sources, some foundries have been continuously adjusting their quotes for mature process downward by single-digit percentages (4% to 6%) this quarter.

With mature process production capacity continuing to expand in China, it is reportedly estimated that prices may decrease further in the second quarter, leading to a cumulative reduction of around 10% for the first half of the year.

Overview of Foundries’ Price Reductions in Mature Process

According to the same report citing sources, IC design companies that previously tape out to Chinese foundries primarily focused on driver ICs. However, recently, some power management IC companies have gradually increased orders to Chinese foundries. Currently, the price difference between foundry services in Taiwan and China can be as high as 20% to 30%.

Regarding this wave of price reductions in mature process, an unnamed sources from the IC design industry cited by the report disclosed that the price reduction for Taiwanese foundries is at least in the low single-digit percentage range (1% to 3%), while for Chinese foundries, it is in the mid-single-digit percentage range (4% to 6%). If the order volume is large, prices can be negotiated even lower, or different discount methods may be available.

Reportedly, as Chinese foundries continue to increase their production capacity for mature process, supported by subsidies from the Chinese government, they maintain considerable flexibility in pricing strategies. As long as customers are willing to provide a certain quantity of orders, prices above the variable costs can be negotiated. Therefore, there is indeed room for further price reductions in the second quarter.

Regarding mature process, IC design companies cited in the report mention that the high-demand 28-nanometer process still faces supply shortages and may even see price increases. However, for the 40-nanometer and 55-nanometer processes, where the increase in production capacity outpaces the return of demand, price reductions are essentially the only option.

With China’s significant investment in mature nodes, it is positioned at a time when the global chip industry is poised for recovery. According to a recent TrendForce’s data, China currently has 44 operational fabs, with an additional 22 under construction. By the end of 2024, 32 Chinese fabs will expand their capacity for 28-nanometer and older mature chips.

TrendForce predicts that by 2027, China’s share of mature process capacity in the global market will increase from 31% in 2023 to 39%, with further growth potential if equipment procurement progresses smoothly.

However, compared to the proactive pricing strategies adopted by Chinese foundries, Taiwanese foundries have been relatively firm in their pricing.

For example, TSMC emphasizes that while it has faced significant pricing pressure from counterparts in China recently, the company is not prepared to engage in a price war. Instead, it anticipates seizing opportunities to attract orders from European and American clients. TSMC aims for moderate growth this year.

Regarding IC pricing, IC design companies cited by the report acknowledge that although their IC production costs have decreased, customers also demand price reductions. Caught between customers and foundries, both of which are larger in scale than IC design companies, it is difficult to predict how well operations will perform this year amidst ongoing challenges.

Read more

• [News] China’s ICs Imports Decrease in 2023, Chinese Manufacturers Focus on Mature Processes

(Photo credit: TSMC)

Reportedly, South Korean memory giant SK Hynix has capitalized on the surging demand in the artificial intelligence (AI) and high-performance computing (HPC) markets. As per a report by ComputerBase, with its HBM and DDR5 products, SK Hynix has swiftly emerged from the slump in the memory market in 2023 and anticipates further growth. Consequently, a new phase of expansion is underway.

The report further indicates that SK Hynix plans to invest at least KRW 120 trillion (approximately USD 89.4 billion) to construct a new semiconductor production complex in Yongin, located in the central part of Gyeonggi Province, South Korea. This includes four independent fabs, with preparations currently underway, one-third of which has already been completed.

The report indicates that SK Hynix announced plans to build the world’s largest chip production facility as early as 2019. However, due to various reasons, the project was delayed until 2022 when agreements were reached with the central and local governments of South Korea, allowing the project to progress.

SK Hynix intends to commence its expansion project officially in March 2025, with the first fab scheduled for completion in 2027 and the entire complex expected to be completed by 2046. It is yet clear whether the first fab will produce DRAM or NAND Flash memory. However, given the significant demand for HBM products in the AI market, and considering SK Hynix’s tight production capacity, this is likely the direction they will choose.

HBM, a type of DRAM primarily used in AI servers, is experiencing a surge in demand worldwide, led by NVIDIA. Moreover, according to a previous TrendForce press release, the three major original HBM manufacturers held market shares as follows in 2023: SK Hynix and Samsung were both around 46-49%, while Micron stood at roughly 4-6%.

Additionally, the four planned fabs are expected to occupy half of the complex’s size, with SK Hynix also constructing numerous supporting facilities in the area, such as wastewater treatment plants and resource recycling centers. Apart from SK Hynix, Samsung has also opted to construct a similar semiconductor production complex nearby, which includes research and development centers, to meet the anticipated market demands ahead.

Read more

• [News] SK Hynix Reportedly Invests USD 1 Billion in Advanced Packaging to Strengthen HBM Chip Leadership
• [News] SK Hynix Rumored to Propose Collaboration with Kioxia for HBM Production in Japan

(Photo credit: SK Hynix)