Samsung Electronics Co. has lost a competition with Taiwan Semiconductor Manufacturing Company (TSMC) for the production of graphics processing units (GPUs) for Nvidia.
Taiwanese media reports celebrate winning the competition but the market rather takes a positive view on Samsung Electronics. Samsung was defeated by TSMC, the world’s largest dedicated semiconductor foundry, this time as the latest deal is to produce a large-sized GPU. However, the next round the battle is likely to be determined by the ability to reduce the chip size.
According to industry sources and foreign media reports on Sept. 19, TSMC will exclusively produce a GPU to be used in Nvidia’s newest flagship graphics card “GTX 2080 Ti.” Initially, Nvidia was said to use Samsung’s 10-nanometer (nm) GPU in its new product, but Nvidia finally chose TSMC’s 12-nm GPU.
The industry believes that Nvidia has made such a decision considering the stability in supply and unit cost. When Nvidia uses a 12-nm GPU, the size of the chip gets bigger. However, the company put a stable yield rate first. Samsung Electronics’ 10-nm GPU has a reasonable price but it is still more costly than a 12-nm GPU.
Chinese media reported that it is TSMC’s victory over Samsung. TSMC has been seeking to keep Samsung Electronics in check as the Korean electronics giant has pledged to become the world’s second largest foundry firm. TSMC posted US$32.2 billion (36.13 trillion won) in sales and had a 50.4 percent share in the global foundry market last year. Samsung’s capability still falls short of TSMC. Samsung spun off its foundry operation from the System LSI Division last year and has been making a massive investment, but its market share stands at only 6.72 percent. A Taiwanese media source pointed out, “Nvidia’s latest decision must have put more pressure on Samsung.”
However, some say that it is early to regard the outcome of the competition simply as Samsung’s defeat. Nvidia’s latest graphics card series to be launched in two years is a large GPU, like a server central processing unit (CPU). Therefore, the company has to reduce the size for the next new product. This is why Samsung Electronics is forecast to receive an order to produce a 7-nm to 8-nm GPU for Nvidia’s next series after giving its 10-nm GPU a miss.
TSMC is also ahead of Samsung in the 7-nm foundry business but this is highly likely to become a huge opportunity for Samsung. TSMC succeeded in 7-nm process test mass production at the end of last year and it already secured the mass production system in the second quarter this year. Samsung Electronics has also set up its 7-nm chip mass production system. Samsung’s prospect has improved as Global Foundries (GF), which is now the world's second largest foundry business, has stopped 7-nm process development.
Samsung Electronics aims to outpace TSMC in the nano fabrication process sector based on its production lines using extreme ultraviolet (EUV) lithography equipment. When Samsung Electronics wins a large amount of orders for not only graphics cards but also artificial intelligence (AI) chip production from market leader Nvidia, its status in the foundry industry is expected to greatly improve.