Intel has chosen TSMC as a manufacturer for their high-performance graphics processing unit (GPU). It seems that TSMC is actively utilized during the transition period before securing an ultra-fine process foundry. TSMC is also known to prepare for the production of server central processing units (CPUs) and GPUs applied with the Intel 3 nanometer (nm) process, and the cooperative system between the two companies is becoming stronger.

Photo Image
<Intel Senior Vice President Raja Koduri introduced the new GPU brand Arc at Intel Architecture Day.>

At Architecture Day event that presents the next-generation architecture roadmap on the 19th (local time), Intel announced that they plan to produce the next-generation supercomputer GPU 'Ponte Vecchio' and PC gaming GPU 'Arc' at TSMC. Ponte Vecchio will use TSMC's 7 nano and 5 nano processes and 6 nano process for Arc production.

Earlier this month, Intel announced plans to release the Ponte Vecchio and announced that some GPU chipsets would be produced using an external foundry. As a result, Samsung Electronics and TSMC were in the spotlight. This is because Samsung Electronics and TSMC are the only foundries capable of micro processing less than 7 nm that can produce next-generation GPUs.
 
At Architecture Day, as Intel selected TSMC as its GPU production base, Samsung Electronics suffered a hardship. The prevailing opinion is that Intel selected TSMC, which has superior foundry competitiveness than Samsung Electronics, after reviewing various aspects. According to market research firm TrendForce, TSMC's foundry market share was 55% and Samsung Electronics' was 17% as of the first quarter. Some analysts say that it is a countermeasure aimed at Samsung Electronics, which is easier to follow than TSMC with strong market share.

According to Taiwanese media, Intel also has entrusted TSMC for production of server CPUs and GPUs with 3 nano process. It is known that the production line will be operated from February next year, and Intel products will be mass-produced from July. As reported, Intel will produce CPUs and GPUs using a number of ultra-fine processes from TSMC's 3 nano to 7 nano.

Intel's move is interpreted as a strategy to actively utilize TSMC until their own foundry is stabilized. Intel announced that it would re-enter the foundry market earlier this year, but it will take some time to start mass-production of 7 nano or less processes. Full-scale mass production of the 'Intel 4' process, which is estimated to be similar to the industry's 5 nano process, is scheduled for the second half of next year. It is interpreted as Intel will to utilize TSMC as a manufacturer of major products until Intel foundry starts in earnest.

An industry official said, "Intel will produce most of its flagship products, including CPUs on their own, and expand their foundry in only some areas, but until their production capacity is secured, they will inevitably depend on consignment production to some extent. Therefore, there is tendency of the increasing proportion of TSMC in consignment production.”

By Staff Reporter Dong-jun Kwon (djkwon@etnews.com)