Quantcast
Channel: EDN
Viewing all articles
Browse latest Browse all 655

The 2-nm process node and Samsung’s foundry crossroads

$
0
0

Despite being the first to adopt the gate-all-around (GAA) technology, Samsung has struggled with commercialization and yield issues and its foundry market share has dropped below 10%. Han Jin-Man, recently named president of Samsung’s foundry operations, aims to change that with a major strategic shift in the execution and production of 2-nm manufacturing node.

Previously, Han oversaw the DRAM, flash memory design, and SSD development teams at Samsung. The Korean chipmaker has also named process development specialist Nam Seok-Woo as the chief technology officer (CTO) of the foundry division. It’s a major reshuffle at the world’s second-largest chip contract manufacturer, currently trying to close the gap with market leader TSMC, which has a market share of nearly 65%.

Figure 1 Like Intel, there is now an industry chat about the potential spinoff of Samsung’s foundry business. Source: Samsung

The change of guard at Samsung’s foundry business comes as it transitions to the next-generation foundry process: the 2 nm chip manufacturing process. The shift to the 2-nm process node entails a significant overhaul of transistor structures, which inevitably leads to higher development costs and greater design complexity.

While Samsung’s foundry business has never been in a position to challenge TSMC, it stunned the market by surpassing TSMC in launching the GAA technology to replace the FinFET manufacturing technique. While this significantly raised Samsung’s foundry profile, commercializing this highly complex technology proved a tough nut to crack.

The GAA setback

In 2022, Samsung made waves by incorporating GAA transistors at its 3-nm process node; meanwhile, TSMC decided to implement GAA technology at its upcoming 2-nm process node. In retrospect, that seems to be the right call because Samsung faced difficulties achieving high yields due to GAA’s technical complexity.

In GAA technology, transistors use vertically placed horizontal nanosheets, enabling the gate to cover the channel on all four sides. That reduces current leakage and improves drive current, resulting in chips with enhanced performance and better energy efficiency.

Figure 2 While a gate covers three sides in the FinFET structure, in GAA, a gate surrounds the entire four sides of a cylindrical channel, which is a passage for current flows. Source: Samsung

Now, Samsung is determined to recover from the setbacks it experienced at the 3-nm process node. It begins with improving yield and better handling GAA’s technical complexity. “We will focus on dramatically improving the yield of the 2-nm manufacturing process,” said Han Jin-Man, Samsung’s new foundry chief.

It’s important to note that TSMC is expected to achieve over 60% yield in trial production and enter mass production for 2nm chips ahead of Samsung. On the other hand, Samsung’s 2-nm yield has been reported in the 20% to 30% area.

Moreover, according to industry sources, TSMC plans to start trial production of its 2-nm process node in April 2025 and mass production in the second half of 2025. On the other hand, Samsung is expected to start test production in the first half of 2025 and mass production in the fourth quarter of 2025.

High time at Samsung

Samsung, already under immense pressure after ceding leadership of high-bandwidth memory (HBM) products to SK hynix, is now in a renewed spotlight for its issues regarding the semiconductor contract manufacturing business. Intel’s troubles in this area have led to the ouster of its CEO and an uncertain future.

The second-best stature doesn’t often spark envy in the semiconductor industry, and Samsung isn’t known to be shy about fighting back. Its quest for relevance in the 2-nm space has already been rewarded with a manufacturing order from Japanese AI chip startup Preferred Networks (PFN), whose investors include Toyota, NTT, and FANUC.

Besides manufacturing 2-nm chips, Samsung will also provide its 2.5D packaging technology—I-Cube S—to enable PFN to integrate multiple chips into a single package. That marks a significant breakthrough, and if Samsung can secure one major customer of TSMC, it’ll create a significant impact in the contract foundry business.

Still, catching up with TSMC remains a daunting challenge.

Related Content

The post The 2-nm process node and Samsung’s foundry crossroads appeared first on EDN.


Viewing all articles
Browse latest Browse all 655

Trending Articles