Pioneering Offline AI: Semifive and ICY Tech Unveil Advanced Semiconductors for Smart Devices
Seoul, Thursday 7 May 2026
SEMIFIVE and ICY Tech have developed an advanced semiconductor that remarkably enables complex AI models to function entirely offline within smart devices and humanoid robots across Asia.
The Mechanics of Edge AI and eMRAM
On 6 May 2026, the custom semiconductor provider SEMIFIVE and the Chinese artificial intelligence firm ICY Tech announced the successful tape-out of a next-generation Edge AI System-on-Chip (SoC) [1]. This collaboration leverages Samsung Foundry’s 8-nanometre embedded Magnetoresistive Random Access Memory (eMRAM) technology, marking the first commercial deployment of this specific node in Asia [1]. The architecture is explicitly designed to support on-device inference for models containing up to 2 billion parameters, allowing devices such as AI personal computers and humanoid robots to operate efficiently in offline environments [1]. eMRAM itself represents a critical advancement in memory architecture; it combines the non-volatile data retention of NAND flash memory with the operational speed of DRAM, positioning it as a highly efficient successor to traditional embedded flash memory [1].
Surging Global Demand and the AI Infrastructure Boom
The successful tape-out by SEMIFIVE and ICY Tech coincides with a period of explosive growth across the broader semiconductor sector. According to the Semiconductor Industry Association (SIA), global semiconductor sales reached $298.5 billion in the first quarter of 2026 [3]. This represents a robust sequential increase of 25 per cent compared to the previous quarter, a highly irregular and notable achievement given that the first quarter is traditionally the weakest period of the year [3]. John Neuffer, the CEO of the SIA, noted that the industry remains on a clear trajectory to achieve $1 trillion in global sales by the end of 2026 [3]. This macroeconomic tailwind is heavily driven by the demand for artificial intelligence infrastructure, a trend clearly reflected in the financial performance of major architectural designers.
The Critical Role of Advanced Materials
Beneath the high-level architectural designs and billion-dollar revenue projections lies the foundational chemistry that makes advanced semiconductor manufacturing possible. The proliferation of AI chips and high-performance devices has driven a parallel surge in demand for critical consumables, particularly Chemical Mechanical Planarisation (CMP) slurries [4]. These slurries are essential for polishing silicon wafers and other microelectronic surfaces to the extreme atomic flatness required for modern chipmaking [4]. In 2025, the global CMP slurries market was valued at $1,867.3 million [4]. Driven by the complexities of advanced packaging and hybrid bonding technologies used in AI hardware, this market is projected to reach $2,998.47 million by 2032 [4]. This represents an absolute growth of 60.578 per cent over the seven-year period, corresponding to a compound annual growth rate of 7 per cent [4].