Samsung is reportedly mulling to update its fabrication capabilities in the Taylor fab near Austin from 4nm to 2nm, and delaying the related equipment orders, according to reports from Tom’s Hardware and Korea media outlet ETnews. The decision is said to be made in Q3 2024.

Samsung’s initial projection, announced in 2021 when the investment plan was disclosed, had anticipated the Taylor plant to start its mass production of 4nm in the second half of 2024. But afterwards, the tech giant has reportedly postponed the schedule to 2025.

Now, according to the latest reports, with the AI hardware race intensifying and Nvidia, one of Samsung’s biggest clients, keeps advancing in its GPU platforms, 4nm production may no longer suffice. Therefore, Samsung has postponed its equipment orders for the Taylor fab as it considers the upgrade from 4nm to 2nm. The plant may not begin operations until 2026, US local media MySA noted.

According to information from Samsung’s press release, the company plans to provide one-stop solution for clients to expedite their production of AI chips, while its updated 2nm node with backside power delivery is expected to enter the market in 2027. SF4U, its high-value 4nm variant that offers PPA improvements by incorporating optical shrink, plans to enter mass production in 2025.

Earlier in April, the U.S. government announced that it will provide up to USD 6.4 billion in subsidies to Samsung for expanding advanced chip production capacity at its Texas plant, while it previously approved subsidies of up to USD 8.5 billion for US chip giant Intel and USD 6.6 billion for TSMC, respectively.

Regarding major semiconductor companies’ capacity expansion plans in the U.S., TSMC’s new fab in Arizona aims to complete all preparations for mass production by the end of this year, and commences mass production using the 4-nanometer process in the first half of 2025. On the other hand, TSMC also plans to build a third fab in Arizona, which will use 2nm process or even more advanced technologies to manufacture wafers for customers, though the date for construction has yet to be disclosed.

Meanwhile, Intel plans to mass-produce its 20A and 18A nodes at its Arizona and Ohio plants in 2024 and 2025, respectively.

(Photo credit: Samsung)

Following the announcements of AI PC processor chips by NVIDIA and Intel, AMD has also entered the fray by unveiling the Ryzen Pro 8040 and Ryzen Pro 8000 series chips. According to a report from Commercial Times, they will be manufactured by TSMC and are expected to be released in the second half of the year.

On April 16th, AMD announced that the Ryzen Pro 8040 series chip, designed for laptops, and the Ryzen Pro 8000 series chip, designed for desktops, are the most advanced commercial PC chips ever created. These chips will be manufactured using 4-nanometer technology. It is expected that new AI PCs from HP and Lenovo, starting in the second half of this year, will incorporate these two major chip series.

“AI PC” are laptops and desktops capable of directly running real-time language translation and other AI applications locally, as opposed to most PC devices on the market that rely on cloud platforms for AI computations.

Intel previously announced the Core Ultra series chips designed specifically for AI PCs last year and claimed that the first batch of over 230 AI PCs globally will feature this series of chips. Collaborating partners include major PC manufacturers such as Acer, ASUS, Dell, HP, and Lenovo. Gelsinger once expressed at a New York launch event in 2023 that they anticipate the AI PC to be the standout performer in the coming year.

NVIDIA also unveiled a new generation of AI chips in January of this year, claiming they can execute AI applications directly on PC. Collaborating partners include Acer, Dell, and Lenovo.

Similarly targeting the AI PC market, AMD previously announced the Ryzen 8000G series chips designed specifically for desktops in January. Previously, during an interview with Chinese media Sina, Lisa Su, CEO of AMD, indicated that AI-powered PCs will play a crucial role in driving the growth of the PC market this year.

As NVIDIA, Intel, and AMD vie for opportunities in the AI market, TSMC, which possesses the world’s most advanced chip manufacturing technology, emerges as the primary beneficiary. TSMC currently manufactures AI chips for NVIDIA using a 3-nanometer process and is expected to begin mass production of the next-generation 2-nanometer process starting from next year.

Read more

• [News] China Imports 32 Lithography Systems in 2 Months, Shanghai and Beijing Top Buyers
• [News] AMD Unveils Ryzen 8040 Series AI Processor Utilizing TSMC’s Advanced 4nm Process

(Photo credit: AMD)

According to wccftech, Intel’s new GPUs will come in two models, namely Battlemage-G10 (abbreviated as BMG-G10) and Battlemage-G21 (abbreviated as BMG-G21).

These two new GPUs from Intel were revealed in an internal document. According to the document, the BMG-G10, targeted at enthusiasts, is a GPU with a TDP of less than 225W, while the BMG-G21 is designed as a mid-range performance product with a maximum TDP not exceeding 150W.

As for specific parameters and performance, the enthusiast-grade BMG-G10 is expected to be equipped with up to 64 Xe2 cores, directly competing with NVIDIA’s RTX 4070. On the other hand, the mid-range BMG-G21 aims at the RTX 4060, both continuing to utilize TSMC’s 4nm manufacturing process.

Therefore, previous rumors suggesting that Intel had canceled the development of BMG-G10 and only retained the BMG-G21 with 40 Xe2 cores appear to be untrue. Moreover, the core count of BMG-G10 is larger than initially reported at 56 Xe2 cores, indicating it is poised to deliver even higher performance.

Recently, per a report from Reuters, Intel, Qualcomm, Google, and other major tech companies are teaming up to challenge NVIDIA’s market dominance and make inroads into the AI software sector. They are expected to look to steer developers away from NVIDIA’s CUDA software platform, a parallel computing platform tailored for GPU acceleration.

Read more

• [News] Addressing AI PC Chip Solutions, Intel and Qualcomm Reportedly Unveil Respective Strategies
• [News] Intel, Qualcomm, Google Reportedly Form Alliance, Posing Challenges to NVIDIA’s Dominance in AI?

(Photo credit: Intel)

The annual AI event, NVIDIA GTC (GPU Technology Conference), is set to take place on March 17th, as H200 and the next-generation B100 will reportedly be announced ahead of schedule to seize the market. According to Commercial Times’ report citing sources, H200 and the upcoming B100 will adopt TSMC’s 4-nanometer and 3-nanometer processes respectively. H200 is expected to be launched in the second quarter, while it’s rumored that orders for the B100 adopting Chiplet architecture have already been placed for production.

Sources cited by the report also indicate that NVIDIA’s orders are robust, pushing TSMC’s 3 and 4-nanometer production capacity to near full utilization, making the first quarter, traditionally a slow season, unexpectedly busy.

Regarding the matter of NVIDIA’s next-generation chip orders overwhelming TSMC’s advanced processes, TSMC stated that details regarding production capacity remain consistent with the previous earnings call and will not be elaborated further.

Still, Commercial Times further cited industry sources, revealing that TSMC, in response to anticipated capacity constraints by 2023, is accelerating its efforts. Particularly focusing on advanced packaging like CoWoS, they’ve not only relocated equipment from the Longtan facility but also swiftly activated the AP6 plant in Zhunan.

Another industry sources reportedly indicate that the planned construction of the Tongluo
facility, initially slated for the second half of this year, is now scheduled to commence in the second quarter. The aim is to ramp up 3D Fabric capacity to produce 110,000 12-inch wafers per month by the first half of 2027.

Meanwhile, TSMC’s advanced processes remain fully utilized, with capacity utilization exceeding 90% in February, driven by sustained AI demand.

NVIDIA, on the other hand, recently emphasized that computational-intensive tasks like Generative AI and large language models require multiple GPUs. From customer purchase to model deployment, it takes several quarters. Thus, this year’s inference applications stem from GPU purchases made last year. As model parameters grow, GPU demand is expected to expand.

In addition to increasing GPU quantities, NVIDIA’s GPU efficiency is poised for a significant boost this year. The Blackwell series, notably the B100, is hailed as NVIDIA’s next-generation GPU powerhouse by the market.

Not only is it the first to adopt TSMC’s 3-nanometer process, but it’s also the pioneer in Chiplet and CoWoS-L packaging among NVIDIA products. This tackles high power consumption and cooling issues, with projected single-card efficiency and transistor density expected to surpass AMD’s MI300 series set to debut in the first quarter.

Read more

• [News] NVIDIA Expands Advanced Packaging Supply Chain with Intel’s Inclusion, Diverting Orders from TSMC
• [News] NVIDIA CEO Jensen Huang Meets with TSMC Executives in Taiwan, Addressing the Significant AI Chip Shortage

(Photo credit: Kioxia)

According to TechNews’ report, Apple, NVIDIA, AMD, Qualcomm, and MediaTek all utilize TSMC’s semiconductor processes for manufacturing their latest chips, with some potentially employing Samsung’s foundry, though typically not for flagship products.

With Samsung’s improved yield rates in recent months, the company is eager to secure a portion of the orders, particularly for the 3-nanometer GAA (Gate-All-Around) process.

Earlier market reports suggested that Qualcomm’s Snapdragon 8 Gen 4 might adopt a dual-foundry strategy, simultaneously utilizing TSMC’s N3E process technology and Samsung’s SF3E process technology.

However, both Qualcomm and MediaTek currently plan to employ TSMC’s second-generation 3-nanometer process technology (N3E) for manufacturing chips like the Snapdragon 8 Gen 4 and Dimensity 4, without pursuing a dual-foundry strategy at this time.

As of the end of June 2022, Samsung announced the commencement of production for 3-nanometer process chips at its Hwaseong Industrial Complex in South Korea. These chips incorporate a new GAA transistor architecture technology, rumored to be more energy-efficient compared to TSMC’s 3-nanometer FinFET technology. Despite this, in the realm of 3nm, Samsung has yet to secure substantial orders from major clients.

Interestingly, the company has seen more success in the 4nm domain. It is reported that Samsung has gradually addressed yield and various issues in the 4-nanometer process technology domain. The third generation of 4-nanometer process technology has seen improvements in performance, reduced power consumption, increased density, and achieved yields close to TSMC’s level. Market sources indicate that Samsung has gained recognition from companies like AMD and Tesla, securing new orders.

Currently, TSMC’s 3-nanometer process technology production capacity is ramping up, with an expected monthly capacity of 100,000 wafers by the end of 2024. The revenue contribution is projected to increase from the current 5% to 10%.

Meanwhile, Samsung plans to introduce the second generation of its 3-nanometer process technology, named SF3 (3GAP), in 2024. Building upon the existing SF3E, it aims for further optimization, and Samsung’s in-house Exynos 2500 is expected to be one of the first high-performance chips to adopt this new process technology.

Read more

• [News] Samsung Reportedly Lands a 4nm Mega Order – Why is AMD Switching to “Dual Foundry Mode” for Its Next-Gen Chips?
• [News] A battle on 4nm: AMD Teams Up with Samsung, while Google Weighs Supplier Split