Tokyo-based AI chip startup EdgeCortix Inc. has received 3 billion yen (US$20.7 million) in government subsidies to develop energy-efficient AI chiplets.

This funding supports Japan’s push to advance domestic semiconductor technologies with possible defense uses.

The AI chiplets, aimed at real-time robot decision-making, are planned for commercialization in 2027.

EdgeCortix will produce these chips at Taiwan Semiconductor Manufacturing Co.’s Japan facility.

This follows a previous 4 billion yen (US$27.7 million) grant for chips used in next-generation communication.

These chips operate without relying on cloud computing, reducing power use and data costs.

🔗 Source: Bloomberg

🧠 Food for thought

1️⃣ Japan’s dual-use technology push represents a strategic defense shift

EdgeCortix’s funding reflects Japan’s broader national security strategy transformation that has been accelerating since 2023.

The Japanese government passed a major economic security bill specifically targeting critical technologies and supply chains, viewing them as essential to national security rather than just economic assets 1.

This pivot is backed by substantial resources, with Japan approving a record JPY8.7 trillion (USD55.1 billion) defense budget for fiscal year 2025—a significant 9.4% increase year-over-year 2.

EdgeCortix’s development of AI chips with explicit defense applications aligns with the government’s identification of approximately 200 startups for potential defense technology collaboration 3.

The company’s founder directly acknowledged this strategic alignment, noting their technology can be used “from land-based systems to aerial platforms to space” for “national and global security”—language that would have been unusual for Japanese tech companies in previous decades.

2️⃣ AI semiconductor startups face fierce competition despite defense opportunities

EdgeCortix operates in an increasingly crowded field of specialized AI chip developers, with direct competitors including Tenstorrent, DEEPX, and Rebellions—all focused on energy-efficient edge AI processing 4.