[Portfolio Spotlight] SEMIFIVE Secures $52M AI Inference Accelerator Development Contract

September 29, 2026 – SEMIFIVE, a K-ASIC portfolio company specializing in custom AI semiconductor (ASIC) design, announced that it has signed a development contract worth approximately KRW 70.3 billion (US$52 million) with a U.S.-based AI fabless company for a next-generation AI inference accelerator.

• 

The agreement is the largest single contract in SEMIFIVE’s history and marks the company’s first “Spec Hand-off” project in the North American market. The contract value represents more than 40% of SEMIFIVE’s total orders in 2025 (KRW 168.4 billion) and approximately 60% of its new orders secured in the first half of 2026 (KRW 118.9 billion).

SEMIFIVE’s First Spec Hand-off Project in North America

Under the Spec Hand-off model, the customer provides the core performance requirements and specifications, while SEMIFIVE manages the development process from detailed design through packaging, testing, mass production, and software development.

Unlike a conventional turnkey model, in which customers typically complete the semiconductor design before outsourcing manufacturing-related processes, the Spec Hand-off model enables SEMIFIVE to lead the development process from the specification stage through production.

Next-Generation AI Inference Accelerator

The project will incorporate technologies designed to address memory bandwidth, data processing performance, power consumption, and total cost of ownership (TCO) requirements for large-scale AI inference.

The accelerator is planned to incorporate LPDDR6, a next-generation low-power DRAM technology, to reduce memory-related power consumption. It will also use PCIe Gen5 to support high-speed data transfer and minimize bottlenecks during large-scale data processing.

SEMIFIVE also plans to apply its experience developing multiple large-die projects of up to 800 mm² to achieve high computational throughput and operational stability within the chip architecture.

The two companies plan to complete tape-out in the first half of 2027, with global mass production scheduled to begin in 2028 for hyperscalers and cloud service providers (CSPs).

Advancing Custom Silicon for AI Inference

“The rapid growth of the AI inference market is driving increased demand for custom semiconductors optimized for individual services and workloads,” said Myunghyun Cho, CEO of SEMIFIVE. “By successfully commercializing this next-generation inference accelerator, we aim to establish ourselves as an essential custom semiconductor development partner in the global hyperscale infrastructure market.”

---

Disclaimer

This article contains information and materials provided by a K-ASIC portfolio company. K-ASIC does not independently verify all statements, technical specifications, financial information, or forward-looking information included in this article and assumes no responsibility for the accuracy or completeness of such content.