Samsung SAMSUNG's latest industry updates, AI solution empowerment, 3nm chip production, facing delays

Samsung Electronics Co., Ltd., a global leader in advanced semiconductor technology, showcased its latest foundry innovation technology and outlined its vision for the era of artificial intelligence during the annual Samsung Manufacturing Forum (SFF) held at its device solutions headquarters in San Jose, California.

Under the theme of "Empowering the Artificial Intelligence Revolution," Samsung has announced its roadmap for strengthening its process technology, including two new cutting-edge nodes - SF2Z and SF4U, as well as an integrated Samsung artificial intelligence solution platform that leverages its unique advantages in foundry, memory, and advanced packaging (AVP) business.

 

Dr. Cui Shiyong, President and Head of OEM Business at Samsung Electronics, said, "In an era where many technologies revolve around artificial intelligence, the key to achieving artificial intelligence lies in high-performance and low-power semiconductors. In addition to mature GAA processes optimized for artificial intelligence chips, we also plan to launch integrated co packaged optics (CPO) technology to achieve high-speed and low-power data processing, providing our customers with the one-stop artificial intelligence solutions they need to thrive in this changing era."

 

Empowering customer AI solutions with the most advanced process technology roadmap

 

Samsung announced the launch of two new process nodes, SF2Z and SF4U, to strengthen its cutting-edge process technology roadmap.

 

The company's latest 2nm process SF2Z adopts optimized Backside Power Supply Network (BSPDN) technology, which places the power rail on the back of the wafer to eliminate bottlenecks between power and signal lines. Compared with the first generation 2nm node SF2, applying BSPDN technology to SF2Z can not only improve power, performance, and area (PPA), but also significantly reduce voltage drop (IR drop), thereby improving the performance of HPC design. SF2Z is expected to achieve mass production in 2027.

 

On the other hand, SF4U is a high-value 4nm variant that provides PPA improvements by combining optical reduction, with plans to achieve mass production by 2025.

 

Samsung reiterated that the preparation work for SF1.4 (1.4nm) is progressing smoothly, and the performance and yield targets are expected to be achieved in mass production by 2027. Samsung emphasizes its continued commitment to surpassing Moore's Law and is actively shaping future process technologies below 1.4 nanometers through material and structural innovation.

 

Continuously improving GAA maturity

 

With the advent of the artificial intelligence era, structural improvements such as ring gate (GAA) have become a necessary condition to meet power and performance requirements. On SFF, Samsung emphasized the maturity of its GAA technology, which is a key driving factor for empowering artificial intelligence.

 

In its third year of mass production, Samsung's GAA process continues to demonstrate sustained maturity in terms of yield and performance. With accumulated GAA production experience, Samsung plans to mass produce its second generation 3nm process (SF3) in the second half of this year and achieve GAA on the upcoming 2nm process.

 

Since 2022, Samsung's GAA production has been steadily increasing and is expected to expand significantly in the coming years.

 

Samsung Artificial Intelligence Turnkey Solution

 

Another highlight is the release of Samsung's AI solution, a turnkey AI platform that is the result of the joint efforts of the company's OEM, memory, and AVP business departments.

 

By integrating the unique advantages of various businesses, Samsung provides high-performance, low-power, and high bandwidth solutions that can be customized according to specific customer AI needs.

 

Cross company collaboration has also simplified supply chain management (SCM) and shortened product launch time, resulting in a significant 20% increase in total turnaround time (TAT).

 

Samsung plans to launch an integrated, CPO integrated AI solution in 2027, aiming to provide customers with a one-stop AI solution.

 

Diversify customers and applications, achieve a balanced investment portfolio from artificial intelligence to mainstream technology

 

Samsung has also made significant progress in diversifying its customer base and application areas.

 

In the past year, close cooperation with customers has led to an 80% increase in AI sales at Samsung's OEM factories, reflecting its commitment to meeting the ever-changing market demands. Samsung predicts that by 2028, its AI related customer list will expand fivefold and revenue will increase nine times compared to now. The company has announced several new production technologies and the layout of future artificial intelligence related chips, and stated that this will help it win customers.

 

In addition to cutting-edge technology nodes, Samsung also offers professional and 8-inch wafer derivative products, continuously improving PPA and having strong cost competitiveness. With this balanced technology combination, the company can meet the needs of customers in the fields of automotive, medical, wearable devices, and IoT applications.

 

3nm chip production faces delays

 

Under the prosperity triggered by artificial intelligence, large technology companies around the world are vying for 3-nanometer contract manufacturing products, and Taiwan's TSMC has almost monopolized the market. TSMC's 3-nanometer supply is expected to be sold out before 2026.

 

Despite Samsung Electronics attempting to reverse the "TSMC bias" of leading technology giants including NVIDIA through initial mass production of a 3-nanometer process, its strategy still faces numerous difficulties. Therefore, Samsung Electronics will prioritize improving the yield (yield) of the 3-nanometer chips for its upcoming Galaxy devices as its OEM business unit.

 

According to foreign media and semiconductor industry sources on June 11th, globally renowned semiconductor design companies such as Apple, Qualcomm, NVIDIA, AMD, etc. have taken a large number of TSMC 3 nanometer products, and the market generally predicts that there will continue to be a "TSMC 3 nanometer shortage" in 2026. As Intel's new CPUs also adopt TSMC's 3-nanometer process, it is expected that the competition among the four major semiconductor design companies for TSMC's 3-nanometer production capacity will become more intense.

 

Taiwan's United Daily News quoted semiconductor industry insiders as saying, "Although TSMC's 3-nanometer production capacity has tripled compared to last year, there are still shortcomings." The report also stated that "TSMC plans to convert some of its 5-nanometer facilities in the next two years to support 3-nanometer production and ensure stable production and supply."

 

The reason for the shortage of TSMC's 3-nanometer products is the increase in demand for artificial intelligence servers and Apple's upcoming next-generation iPhone. It is worth noting that Apple will launch the iPhone 16 series as early as September, which will include artificial intelligence features. Like previous generations of products, the iPhone 16 series will be exclusively produced by TSMC, making Apple the largest source of revenue for TSMC's 3-nanometer products.

 

The industry believes that TSMC has achieved an absolute victory over Samsung and Intel, as TSMC's 3-nanometer process production was six months ahead of Samsung's, dispelling concerns about the previous crisis. On June 4th, TSMC's new chairman, Andy Lau, announced at the annual shareholders' meeting that TSMC will increase the price of the 3-nanometer process and confidently declared that "TSMC has no competitors.".

 

Samsung Electronics achieved the world's first mass production of a 3-nanometer process in June 2022 and is currently prioritizing increasing production. Especially, Samsung has placed a heavy bet on Galaxy products equipped with 3-nanometer chips. The Galaxy Watch 7 series, which will be launched in July, will use a 3-nanometer Exynos W1000 (code name Sapphire) chip. In the second half of this year, Samsung plans to mass produce the Exynos 2500 chip, a 3-nanometer application processor, for the Galaxy S25 series.

 

According to TrendForce, Samsung's share in the wafer foundry market decreased from 11.3% in the previous quarter to 11% in the first quarter of this year, while TSMC's share increased from 61.2% in the same period to 61.7%.

 

Driven by the demand for artificial intelligence computing system components, the profitability of this Korean chip manufacturer is recovering. This not only enhances the strength of its main memory chip department, but also provides an opportunity for it to win outsourcing orders.

 

But Samsung must prove that its production is advanced and reliable enough to attract demanding customers such as Nvidia Corp. to make greater promises. The AI accelerator produced by Nvidia is essential for all large technology companies. Samsung also faces challenges from Intel, which is opening factories in an attempt to win orders from its former competitors.

 

The progress of production technology is usually marked by the decreasing size of transistors, which helps to improve the performance of electronic components. Competing to reduce size is the key to winning orders for artificial intelligence processors, which are currently one of the highest performing and most expensive chips in use.