SEMIFIVE has begun mass production of the ‘Bertha’ AI inference accelerator for HyperAccel, a data center chip exceeding 500 mm² in size and built on Samsung’s 4nm process. This achievement represents SEMIFIVE’s third mass-production AI chip program within a year, following projects for Hanwha Vision and a Japanese customer, and signals a shift towards sustained revenue, the company says.

“As demand for ASIC accelerators in data centers is surging with the shift from AI training to inference,” said the CEO of SEMIFIVE. The company reports securing KRW 42.3 billion in new mass-production orders in the first half of this year.

Bertha AI Inference Accelerator Launches on Samsung 4nm Process

SEMIFIVE successfully provided a complete solution encompassing design, verification, packaging, testing, and volume manufacturing. This achievement signifies a full-scale expansion for the company, building on previous mass-production projects for Hanwha Vision’s ‘Wisenet 9’ in the third quarter of last year and a high-performance computing chip for a Japanese client earlier this year, according to SEMIFIVE. The accelerator is specifically designed to optimize artificial intelligence inference workloads, including large language models.

This launch marks SEMIFIVE’s first large-scale project using Samsung Foundry’s 4nm technology, demonstrating a shift towards recurring revenue streams beyond one-time projects. The company’s ability to secure and deliver on these consecutive mass-production programs, three within a year, highlights a strengthening business model. Financial growth is already evident; SEMIFIVE reported KRW 42.

Quarterly order intake increased by 71%, from KRW 15.6 billion in the first quarter to KRW 26.7 billion in the second, with 45% of second-quarter bookings originating from international clients. The company intends to maintain this momentum throughout the remainder of the year, supported by secured end customers and high-value data center ASIC projects, establishing a positive cycle of design wins translating into revenue.