Samsung's new US chip fab wins first foundry order from Groq

Samsung Electronics Co. has bagged the first foundry order for its new factory that will be up and running later next year in Taylor, Texas, from American fabless startup Groq Inc., a move expected to further heat up foundry competition with its bigger rival Taiwan Semiconductor Manufacturing Co. (TSMC) in the US.

Groq on Sunday announced that it has chosen the South Korean chip giant as its “next-gen silicon partner,” and the chips, its new AI chips, will be manufactured on the 4-nanometer (nm) process, or the SF4X process, at Samsung’s new chip factory under construction in Taylor, Texas.

This is the first order Samsung’s foundry business has won for its new chip manufacturing facilities in the US, which are scheduled to commence mass production in the second half of 2024.

Upon its opening late next year, the new Samsung chip factory is expected to start producing Groq’s chips immediately, about a year earlier than the anticipated maiden operation of global foundry giant TSMC’s new US chip factory in Phoenix, Arizona.

The Taiwanese foundry company originally planned to begin mass production at its first chip factory in Arizona by late 2024 but has pushed back its opening to 2025 due to a shortage of qualified and skilled workers.

The two foreign foundry majors are aggressively expanding their foundry business in the US after US President Joe Biden last summer signed into law the CHIPS and Science Act, which includes over $52 billion in semiconductor subsidies to boost chip manufacturing in the US.

The Korean memory giant has committed a total $17 billion to construct the new chip factory in Taylor.  

Samsung’s latest win is expected to further intensify the two rivals’ race to win foundry customers in the US, said analysts.

Their rivalry is fierce, especially in advanced processing technology such as the 4 nm and finer 3 nm nodes. Samsung is known to have significantly improved the chip manufacturing yield for the 4 nm node.  

The world’s No. 1 foundry company TSMC and runner-up Samsung have vowed to begin mass production of advanced chips on the industry’s cutting-edge 2 nm technology in 2025.

GROWING AI CHIP MARKET

Groq is a fabless startup headquartered in Mountain View, California. It develops inference AI chips and accelerators for large language models and more, the company explains.

Inference is one of the two main phases of machine learning, along with training. In the training phase, machine learning focuses on learning everything about datasets, and then it makes predictions from live data to produce actionable results during the inference process.

Groq’s ultra-fast language processor unit boasts a two to four times power advantage compared with other AI solutions, the company said.

It also claims that its current generation AI chip offers the lowest latency machine learning architecture on the market.

Its next-generation AI chip, to be manufactured by Samsung, will be jointly designed with Samsung’s Foundry Design Service team, according to Groq.  

Along with Tenstorrent Inc. and Cerebras Systems Inc., Groq is considered a promising startup poised to lead the global AI chip market.

Groq raised $300 million from global venture capitalists, including Tiger Global Management and D1 Capital Partners, in April 2021.

Its CEO and founder Jonathan Ross and a majority of its other executives and employees are former Google AI chip developers and engineers.

Ross expected that Groq’s partnership with Korean chip giant Samsung armed with “the most advanced semiconductor manufacturing technologies available” would help the American fabless startup take a leap forward.

“Samsung Foundry is committed to advancing semiconductor technology and bringing groundbreaking AI, HPC and data center solutions to the market,” March Chisari, the head of Samsung’s US Foundry business and head of Samsung Semiconductor Innovation Center, was also quoted as saying in a statement.

“This relationship with Groq is further proof of how we’re using our advanced silicon manufacturing nodes to bring new AI innovation to market.”

Write to Jeong-Soo Hwang at hjs@hankyung.com


Sookyung Seo edited this article.