Samsung Proposes Plan to Deliver AI Chips Faster

Samsung Electronics’ contract manufacturing arm unveils a plan to expedite production of AI chips. The new service streamlines communication between clients and Samsung’s memory, foundry, and packaging teams, potentially reducing production time by 20%

Samsung announced on Wednesday that the time required to produce AI chips, which typically takes weeks, has been reduced by approximately 20% due to the implementation of a solitary communication channel that coordinates Samsung’s memory chip, foundry, and chip packaging teams.

Siyoung Choi, President and General Manager of Foundry Business, stated at a Samsung event in San Jose, California, “We are truly living in the age of AI; the emergence of generative AI is completely changing the technology landscape.”

Choi stated that Samsung anticipates the global semiconductor industry’s revenue will increase to $778 billion by 2028, with AI chips catalyzing this growth.

Marco Chisari, Executive Vice President of Foundry Sales and Marketing, stated during a press briefing before the event that the company is confident in the validity of OpenAI CEO Sam Altman’s informal projections regarding the increasing demand for AI processors.

Altman has previously disclosed to contract chip manufacturer TSMC executives that he intends to construct approximately thirty-two additional semiconductor factories. Reuters reported this information.

Samsung is one of the few companies that provides foundry services, designs chips, and sells memory processors under the same roof.

In the past, this combination has frequently been detrimental to it, as some clients were concerned that conducting business with its foundry could benefit Samsung as a competitor in a different sector.

Nevertheless, Samsung anticipates that its turnkey approach will serve as an advantage in the future due to the increasing demand for AI processors and the necessity for all chip components to be highly integrated to rapidly train or infer vast amounts of data while consuming minimal power.

The South Korean technology behemoth also promoted its state-of-the-art chip architecture, gate-all-around (GAA), and transistor architecture that enhances chip performance and minimizes power consumption.

GAA is crucial for the continued development of more powerful processors for AI, as the fineness of the chips is approaching the limits of physics.

Samsung intends to mass-produce its second-generation 3-nanometer circuits using GAA in the latter half of this year, even though global foundry No. 1 TSMC is also developing chips using GAA. Samsung began implementing GAA earlier.

Samsung also disclosed its most recent 2-nanometer chipmaking process for high-performance computing processors. This process involves the placement of power rails on the backside of the wafer to enhance power delivery. The anticipated date of mass production is 2027.