Samsung on Wednesday unveiled an updated roadmap for its most advanced chip nodes aimed at AI chips. The South Korean tech giant also introduced a new turnkey service that leverages its multiple chip business areas to attract companies like Nvidia and AMD to use its foundry or contract chip production service for their AI chips.
The announcement marks the shift in focus by Samsung Foundry, the company’s contract chip manufacturing business unit, toward chips for AI and high-performance computing (HPC) rather than processors in mobile devices.
Also: Samsung’s chip division wants to regain first place: What is happening?
Samsung Foundry’s AI sales increased 80% over the past year and it was making significant progress in diversifying its customer base and application areas amid evolving market demand, the company said. The tech giant aims for more than 50% of its foundry revenue to be earned outside the mobile sector, he also stated.
At Samsung Foundry Forum, its annual foundry event themed this year as Empowering the AI Revolution, held in San Jose, the company showed off its new 2 nanometer (nm) and 4 nm process nodes, called SF2Z and SF4U, respectively.
SF2Z incorporates what is called a back power supply network (BSPDN) to a conventional 2nm node (Samsung’s SF2), where the power rails are placed on the back of the wafer. Today’s chips have all the components of a chip, such as memory, logic, and power rails, on the front of the wafer. Other contract chipmakers are also preparing their own BSPDN technologies – Intel calls it PoweVia and TSMC refers to it as Super PowerRail – which they also plan to adopt for their 2nm or smaller node chips.
Also: Samsung to form HVAC joint venture with Lennox to expand sales in North America
According to Samsung, placing the power rails on the back improves power, performance and area (PPA) and voltage drop. SF2Z is aimed at high-performance computing (HPC) and AI chips and will be launched in 2027. Samsung previously said that SF2 will be launched in 2025 ahead of the forum.
Samsung was the first to start production of a 3nm gate-all-around (GAA) node in 2022. At the forum, Samsung noted that its GAA process was maturing in terms of performance and performance and that a second-generation 3nm node called SF3 will be launched later this year. GAA will also be adopted at the launch of its 2nm node next year, the company said.
Samsung is adding several variants to its 2nm node, with a particular focus on AI. Image: Samsung
Meanwhile, SF4U is a variant of its 4nm SF4 process that offers PPA improvements through optical shrinkage, where a customer’s existing die design is shrunk to fit the newer node. This saves them costs as no major architectural changes to the chip design are required for them to migrate to the more advanced node. SF4U will be launched in 2025, while Samsung already offers SF4 to customers. The South Korean tech giant also reiterated that its 1.4nm node (SF1.4) will be launched in 2027 and was also preparing for sub-1.4nm chips.
Samsung also introduced its new turnkey foundry platform called Samsung AI Solutions. This platform is offered jointly by Samsung’s three business units in its chip division, Foundry, Memory and Advanced Package (AVP).
Also: Samsung Galaxy Ring: features, price, availability and everything we know
According to the company, the platform integrates the “unique strengths” of each of these business units that will allow Samsung to offer solutions to customers tailored to the specific requirements of its AI chips. Overall, it will provide more bandwidth in a compact form factor, reduce power consumption and improve signal integrity, Samsung said.
Because it is the only chip maker that can offer memory chips to accompany the customer’s chip, manufacture these chips and package them, Samsung said this streamlines supply chain management and reduces time to market for the customer. The company said Samsung AI Solutions delivers a 20% improvement in total response time for customers. Samsung AI Solutions will be offered with packaged optics (CPO) __ where even the optics are packaged __ in 2027, the South Korean tech giant said.
Samsung’s decision to add more variants to its process nodes makes sense as a strategy for AI, an umbrella term that actually encompasses a wide variety of different chips designed for different tasks and scales. Some are designed for ChatGPT and other large language models. Some are for image and vision processing for use in drones and display devices. Others are AI accelerators for data centers.
According to market research firm Omdia, the global foundry market is expected to grow 18.1% annually on average from 2023, when it was worth $103.55 billion, to 2027, when it is expected to be worth $201.28 billion. Dollars. Growth in 3nm and smaller edge nodes will see the sharpest growth at 92.3% annually on average over the period, the research firm’s forecast noted.
Samsung is the world’s largest memory chip producer and the second largest contract chip manufacturer. Your memory business tries to win Nvidia as a customer of its HBM3E chips, which are used in conjunction with the GPU manufacturer’s AI accelerators.