The Korea Herald


SK’s Sapeon, Rebellions to merge for AI chip push

By Jie Ye-eun

Published : June 12, 2024 - 16:33

    • Link copied

Sapeon CEO Ryu Soo-jung (Newsis) Sapeon CEO Ryu Soo-jung (Newsis)

SK Telecom said Wednesday that it is seeking to merge its fabless unit Sapeon with Korea-based chip designer Rebellions, as part of its renewed push for artificial intelligence chips.

The merger push is the result of an agreement between the two sides to create a “homegrown company” to lead the global AI race, the telecom giant said.

The market for neural processing units powering AI services is growing rapidly, driven by the widespread adoption of AI across industries. Competition is also heating up among global tech firms to secure shares in the lucrative market.

Citing the next two to three years as a “golden time” for Korea to gain an upper edge in the global AI chip market, SK Telecom said it plans to finalize the merger within the third quarter of this year and set up a new entity by the end of the year.

Rebellions is expected to be responsible for managing the merged entity, while SK Telecom remains a strategic investor. Sapeon’s shareholders SK Square and SK hynix also pledged their full support for the new entity’s global expansion.

SK Telecom’s crosstown rival KT Corp., which is also a strategic investor in Rebellions, said it plans to add forces for the successful launch of the new entity.

Sapeon, founded in 2016 as an AI chip spinoff from SK Telecom’s internal research and development organization, debuted an AI chip for data centers, the nation’s first of its kind, in 2020. Last November, it unveiled its next-generation AI chip, X330, to expand its presence in advanced AI chips.

Founded in 2020, Rebellions has launched two AI chips, and its market valuation is estimated at 800 billion won ($581 million). Its latest Atom AI chip is the nation’s first NPU commercialized for data centers, with mass production set to start this year. The company is currently working on a new chip, called Rebel, targeting the massive language model market.