SEMIFIVE has signed a contract worth approximately KRW 70.3 billion (~US$52 million) with a U.S.-based AI fabless company to develop a next-generation AI inference accelerator, the Korean custom silicon specialist announced from Seoul on Sept. 29, 2026.

The numbers put the deal in sharp relief. This single contract exceeds 40% of SEMIFIVE's total 2025 orders of KRW 168.4 billion and roughly 60% of the KRW 118.9 billion in new orders the company booked in the first half of 2026. It is SEMIFIVE's largest contract to date and its first "Spec Hand-off" engagement in the North American market.

The Spec Hand-off model sits at the heart of the announcement. Under it, the customer supplies only key performance requirements and specifications; SEMIFIVE then takes responsibility for the entire development chain — detailed design, software development, packaging, testing and mass production. Conventional turnkey engagements typically see customers complete chip design before outsourcing implementation and manufacturing. SEMIFIVE steps in earlier, leading the project from the requirements stage onward.

That earlier involvement reflects a broader shift in the custom semiconductor market, where specialized ASIC vendors now shape silicon from the specification phase. Google and Amazon, for instance, have developed their own AI chips with semiconductor companies such as Broadcom and Marvell — the commercial template SEMIFIVE is chasing with hyperscalers and cloud service providers.

The technical brief is concrete. The accelerator must deliver the high memory bandwidth and data processing performance that large-scale AI model inference demands while cutting power consumption and total cost of ownership. SEMIFIVE will pair LPDDR6, the next-generation low-power DRAM standard, to reduce memory-related power draw with PCIe Gen5 to minimize bottlenecks in high-volume data transfers. The company also plans an architecture delivering high compute throughput and operational stability, drawing on its track record with multiple Big Die projects of up to 800 mm².

The timeline is fixed: tape-out in the first half of 2027, followed by global mass production in 2028, with hyperscalers and cloud service providers as the target customers.

"As the AI inference market continues to expand rapidly, demand is accelerating for custom silicon optimized for each customer's services and workloads," said Brandon Cho, CEO and Co-founder of SEMIFIVE. "By successfully commercializing this next-generation inference accelerator and improving the economics of large-scale AI models, we aim to establish SEMIFIVE as an essential custom silicon development partner for the global hyperscale infrastructure market."

The company, listed on KOSDAQ under ticker 490470, built its business on platform-based SoC design — configurable domain-specific architectures and pre-validated IP pools intended to cut the time, cost and risk of custom silicon. It operates as a leading Samsung Foundry SAFE DSP partner and positions itself as a one-stop shop for SoC design across data center and AI-enabled IoT applications.

If SEMIFIVE hits its 2028 mass production target, the deal would establish its Spec Hand-off model as a credible entry point into a North American custom AI silicon market currently dominated by Broadcom and Marvell.