Samsung to enhance one-stop service for AI chip production

[Courtesy of Samsung Electronics Co.]

Samsung Electronics Co. on Wednesday unveiled plans to introduce advanced foundry technology by 2027 to enhance its one-stop service for AI chip production from development to manufacturing and assembly.

The move is aimed at leveraging its strengths as an integrated device manufacturer (IDM) to meet the growing demand for AI chips and challenge Taiwan Semiconductor Manufacturing Co. (TSMC), the world‘s largest foundry.

On Wednesday, Samsung shared its semiconductor technology strategy for leading the AI era at the Samsung Foundry Forum (SFF) 2024 held in Silicon Valley, the United States.

Under the theme, “Empowering the AI Revolution,” Samsung announced its reinforced process technology roadmap, including two new cutting-edge nodes — SF2Z and SF4U — as well as its integrated Samsung AI Solutions platform harnessing the unique strengths of its Foundry, Memory, and Advanced Package (AVP) businesses.

Currently, Samsung produces high-performance, low-power AI chip products through its integrated AI solution, which combines system semiconductors, high-bandwidth memory (HBM), and packaging.

This integrated approach reduces the time required for chip development and production by approximately 20 percent compared to traditional processes.

The company also aims to simplify the supply chain for customers by offering high-performance, low-power, and high-bandwidth integrated AI solutions through the cooperation of its three semiconductor business units.

By 2027, Samsung plans to introduce new foundry technology, including the SF2Z process, which applies backside power delivery network (BSPDN) technology to its 2-nanometer (nm) process.

The company had previously announced the start of its 2nm process next year and now plans to incorporate BSPDN technology to improve power and signal line efficiency.

Backside power delivery, a complex technique with no commercial examples yet, places power lines on the backside of the wafer to mitigate bottlenecks in power and signal transmission.

This technology is considered a game-changer, enabling ultra-fine processing. TSMC has announced plans to introduce this technology to its 1.6nm process by late 2026.

Samsung Electronics expects that BSPDN technology will improve power consumption, performance, and area (PPA) compared to existing 2nm processes while significantly reducing voltage drop issues, enhancing high-performance computing design capabilities.

The company also plans to integrate optical component technology for AI solutions by 2027, enabling high-speed data processing with low power consumption.

By 2025, Samsung will begin mass production of the SF4U process, enhancing PPA competitiveness through optical shrink technology. The company is also on track to start mass production of its 1.4nm process by 2027, as initially planned, despite TSMC’s announcement of 1.6nm process mass production by 2026.

Since 2022, Samsung has been mass-producing 3nm processes using Gate-All-Around (GAA) technology, the next-generation transistor technology.

The second-generation 3nm process mass production is set to begin in the latter half of this year, with consistent growth in demand expected to drive significant expansion.

To strengthen its foundry business, Samsung is diversifying its customer and application-specific portfolios. The company has seen over 80 percent growth in AI product orders compared to last year by enhancing customer collaboration in the rapidly growing AI sector.

Samsung is also responding to diverse customer demand by offering a portfolio of processes with improved PPA and price competitiveness in both 8-inch foundry and advanced process.

“At a time when numerous technologies are evolving around AI, the key to its implementation lies in high-performance, low-power semiconductors,” said Choi. Si-young, president and head of foundry business at Samsung Electronics.

“Alongside our proven GAA process optimized for AI chips, we plan to introduce integrated, co-packaged optics (CPO) technology for high-speed, low-power data processing, providing our customers with the one-stop AI solutions they need to thrive in this transformative era,” he said.

The event, in the meantime, featured presentations from industry leaders such as Arm CEO Rene Haas and Groq CEO Jonathan Ross, who took the stage to emphasize the robust partnerships with Samsung in tackling new AI challenges.

Building on the momentum of the SFF event, Samsung will host its annual Samsung Advanced Foundry Ecosystem (SAFE) forum on Thursday.

Themed “AI: Exploring Possibilities and Future,” the forum will serve as a collaborative platform for ecosystem partners to discuss customizable technologies and solutions tailored for AI.

The forum will feature the inaugural Multi-Die Integration (MDI) Alliance workshop, following the launch of the MDI Alliance last year, to discuss 2.5D and 3D IC designs for comprehensive solutions development, according to the company.