Seoul – – September 29, 2026 -- SEMIFIVE has signed a contract worth approximately KRW 70.3 billion ($52 million) with a U.S.-based AI fabless company to develop a next-generation AI inference accelerator, marking the South Korean custom ASIC provider's largest single contract to date.
Deal size dwarfs prior order volumes
The contract alone equals more than 40% of SEMIFIVE's total 2025 orders of KRW 168.4 billion and roughly 60% of the KRW 118.9 billion in new orders booked during the first half of 2026. It also marks the company's first "Spec Hand-off" engagement in North America.
Spec Hand-off model shifts design ownership earlier
Under this model, the customer supplies only performance requirements and specifications, while SEMIFIVE handles detailed design, software development, packaging, testing and mass production. This differs from conventional turnkey arrangements, where customers typically finish chip design before outsourcing implementation and manufacturing. Major technology firms including Google and Amazon have similarly worked with semiconductor partners such as Broadcom and Marvell to develop proprietary AI silicon.
Chip integrates LPDDR6, PCIe Gen5 and large-die architecture
The accelerator will use LPDDR6 low-power DRAM to cut memory-related power draw and PCIe Gen5 to reduce bottlenecks in high-volume data transfers. Drawing on prior Big Die projects of up to 800 mm2, SEMIFIVE plans an architecture aimed at high compute throughput alongside operational stability, targeting large-scale AI model inference with reduced power consumption and total cost of ownership.
Tape-out set for first half of 2027, mass production in 2028
Following tape-out, the two companies plan to begin global mass production in 2028, targeting hyperscalers and cloud service providers.
Brandon Cho, CEO and Co-founder of SEMIFIVE, said rising demand for custom silicon optimized to specific customer workloads is accelerating as the AI inference market expands, and the company aims to become an essential custom silicon partner for global hyperscale infrastructure.