SEMIFIVE Begins Mass Production of AI Accelerator on Samsung 4nm

SEMIFIVE has initiated mass production of HyperAccel's AI inference accelerator, dubbed 'Bertha,' utilizing Samsung Foundry's 4nm process node. This development represents SEMIFIVE's entry into large-scale production for data center AI applications, with expectations for steady volume expansion driven by follow-on orders as HyperAccel rolls out its services.

### Money Play If custom AI semiconductor demand accelerates, watch SEMIFIVE's mass production ramp-up for potential revenue gains.

## Catalyst Analysis: AI Inference Accelerator Production SEMIFIVE reported securing an initial mass production contract for HyperAccel's AI inference accelerator, 'Bertha.' The chip is designed for AI inference workloads, including large language models, and is characterized by a large die size exceeding 500 mm². The company highlighted its capability to deliver a complete turnkey solution from design to manufacturing, addressing the complexities associated with advanced node and large-die chip production.

This mass production milestone follows previous programs for security cameras and high-performance computing AI chips, diversifying SEMIFIVE's revenue stream beyond one-time engineering services. The company reported KRW 42.3 billion in new mass-production orders in the first half of 2026, nearly double the full-year intake of KRW 21.2 billion for 2025. Quarterly order intake surged 71% from KRW 15.6 billion in the first quarter to KRW 26.7 billion in the second quarter. Overseas orders constituted 45% of second-quarter bookings.

## $ASIC+WL Technical Analysis & Key Risk Watch

## Sector Ripple / Impact on Semiconductor Manufacturing Companies involved in advanced semiconductor manufacturing, such as those utilizing Samsung Foundry's nodes, may see increased order flow as demand for AI inference accelerators grows. The successful ramp-up of 'Big Die' chips also indicates advancements in packaging and testing capabilities across the supply chain.