SEMIFIVE completes mass production contract for HyperAccel AI chip, signaling emerging chipmaker scale entry.

SEMIFIVE, a leading design solution provider and pioneer of platform-based custom silicon solutions, has announced the conclusion of a contract with HyperAccel to develop a generative AI chip called Bertha for mass production. The collaboration brings together SEMIFIVE's expertise in SoC platforms and comprehensive ASIC design solutions with HyperAccel's cutting-edge AI semiconductor technology. SEMIFIVE specializes in developing expert SoC design platforms for AI chips and is actively developing its SoC chiplet platform in collaboration with industry-leading partners to respond to customer demand for AI custom silicon.

HyperAccel has developed a LLM Processing Unit (LPU), an AI chip specifically designed for transformer-based Large Language Model inference. The company describes this as the world's first semiconductor LPU tailored for LLM inference, offering low cost, low latency, and domain-specific features. The technology is expected to replace existing high-cost and low-efficiency Graphics Processing Units with performance improvements of up to 2 times and a price/performance ratio that is 19 times better than that of a typical supercomputer.

Bertha will be developed using 4nm process technology, with mass production targeted to commence in the first quarter of 2026. According to Joo-Young Kim, CEO of HyperAccel, "We are delighted to work with SEMIFIVE, a leading provider of SoC platforms and comprehensive ASIC design solutions, for the development of Bertha to be mass-produced." Kim added that the collaboration aims to "offer customers AI semiconductors that provide more cost-effective and power-efficient LLM features than GPU platforms," which will "significantly reduce the operational expenses of data centers and expand our business scope to other industries that require LLMs."