AI Venture TAI Completes Evaluation of Next-Generation AI Semiconductor Test Chip "Sting Ray" ~ Transitioning to Mass Production of Edge AI Chips with Proprietary Architecture ~

Tokyo Artisan Intelligence Co., Ltd. (Headquarters: Yokohama, Kanagawa Prefecture, Representative Director, President, CEO & CTO: Hiroki Nakahara, hereinafter "TAI") is pleased to announce that it has completed the evaluation, including design, manufacturing, and testing, of its test chip (development code name: Sting Ray) for reconfigurable AI semiconductor chips for edge AI systems with its proprietary architecture, and has transitioned to mass production.

This development embodies TAI's cutting-edge semiconductor design technology and architectural concepts. Furthermore, with an eye toward future global expansion and mass production, we have acquired the technical know-how for mass-produced chips through cooperation with Oppstar Japan Co., Ltd. (Headquarters: Yokohama, Kanagawa Prefecture, hereinafter "Oppstar Japan"), the Japanese subsidiary of Oppstar (Headquarters: Penang, Malaysia), which possesses advanced semiconductor chip design technology.

With the completion of this test chip, we will fully launch the roadmap for mass production with a view to the future global market and accelerate the construction of our manufacturing and supply chain systems.

Background of Development

With the widespread adoption of generative AI, the increase in power consumption by data centers and the resulting massive heat generation issues have become global challenges. In conventional systems that rely on general-purpose GPUs, environmental impact is reaching its limits in exchange for improved computational power, making the widespread adoption of "dedicated AI chips" with high power efficiency indispensable for achieving GX (Green Transformation).

To solve these social issues, TAI has been independently driving the development of AI semiconductor chips with a proprietary architecture that balances low power consumption and high-speed processing. The completion of this test chip design is a major achievement demonstrating our advanced semiconductor technology and a significant milestone toward bringing next-generation AI chips to market.

Furthermore, to build a supply chain for the upcoming mass production phase, we are pursuing strategic partnerships with external partners who have global manufacturing support capabilities.

About the Test Chip "Sting Ray"

"Sting Ray" is a test chip for technical verification prior to the design of mass-produced chips, developed to solve the critical issue of "increased power consumption due to simultaneous execution of multiple AIs" in the widespread adoption of "Physical AI."

While leveraging the strength of "reprogrammable flexibility" of FPGAs, it adopts the "UMC 40nm process," which enables practical semiconductor manufacturing at a limited cost for a venture company. The main features of this chip are as follows:

1. Four Technical Features

* Verification of flexible "reconfigurability": We are exploring and optimizing a structure that allows flexible modification of processing circuits according to the application and the AI model to be executed. * Optimization of wiring channels: This allows verification of the "wiring channels" that are the core of the architecture, and we have implemented a simple yet extremely efficient structure that allows direct observation and control of wiring status. * Pursuit of low power consumption and low latency: We are verifying the foundation for real-time processing that requires millisecond-level decisions and operates with high efficiency even with limited power resources in the field. * Development of design and verification software: We have developed design software to realize user circuits on reconfigurable semiconductor chips, and verification software to confirm that the manufactured test chips operate correctly.

2. Differences from Conventional Chips (e.g., General-Purpose GPUs)

Conventional general-purpose GPUs and fixed-type AI chips have issues of enormous power consumption and heat generation in exchange for advanced processing, and they can only be optimized for specific AI models. The architecture that "Sting Ray" aims for differs from conventional chips in its ability to simultaneously and flexibly execute multiple AIs according to field requirements while maintaining "low power consumption and low latency."

3. Intended Applications (Physical AI Field)

We envision incorporation into the following Physical AI applications that require advanced on-site judgment:

* Infrastructure/Railways: Inspection and maintenance systems that simultaneously operate numerous cameras and sensors to detect anomalies in real-time. * Manufacturing/Factories: Automation systems that simultaneously perform quality checks and visual inspections on multiple production lines. * Robotics: Industrial and mobile robots that make instantaneous autonomous judgments and control based on changing situations.

Through this project, by independently experiencing the entire process from design to manufacturing, operation confirmation with evaluation boards, and establishment of control software, we have established a basic verification environment for mass production. The knowledge and challenges gained here will be applied to the development of the next-generation mass-produced chip, "Manta Ray."

Future Outlook and Roadmap

Starting with the completion of the "Sting Ray" design, TAI will fully commence exploration of verification, evaluation, and application areas. Moving forward, we will promote the "Manta Ray" project for mass production of next-generation AI semiconductor chips and the construction of a supply chain through collaboration with domestic and international partners.

Specific Initiatives

1. Launch of the "Manta Ray" Project for Mass Production

For the mass production of next-generation AI semiconductor chips, we have decided to adopt the UMC 40nm process. This will enable even startups and venture companies to manufacture and sell AI semiconductor chips at a limited cost. We plan to release the mass production and development environment according to the following roadmap:

* 1Q 2027: Design Software (Alpha Version) * 2Q 2027: Engineering Sample (ES) Chip Manufacturing * 3Q 2027: Evaluation Board Equipped with ES Chip * 4Q 2027: Mass Production (MP) Chip Manufacturing * 1Q 2028: Evaluation Board Equipped with MP Chip

2. Expansion into the "Physical AI" Field Leveraging Low Power Consumption and Low Latency

The "low power consumption" and "low latency" AI processing capabilities of this chip will demonstrate their greatest strengths in field applications (Physical AI field) where real-time performance and reliability are strictly required. Moving forward, we will promote introduction and social implementation in the following specific industrial fields:

* Railways/Infrastructure: Efficiency improvement and safety enhancement through incorporation into inspection and maintenance support systems. * Manufacturing/Inspection: Automation of quality checks and real-time anomaly detection in manufacturing lines. * Robots: Instantaneous situation judgment and control in the field for autonomous mobile robots, industrial arms, etc.

3. Expansion and Communication of Supply Chain Collaboration

With the completion of the "Sting Ray" test chip as an opportunity, we will proceed step-by-step from establishing the development environment to exploring application areas, collaborating with partners, and achieving social implementation.

We will verify the knowledge gained from R&D for industrial applications and explore collaboration possibilities in each phase of evaluation, design, and implementation with domestic and international partner companies and overseas vendors. We will communicate our "low power consumption AI technology," which can contribute to global social issues such as GX (Green Transformation) and labor saving, to the world, and will gradually announce the results as verification progresses.

Furthermore, we plan to showcase a live demo of the test chip announced this time at an event scheduled to be held on August 6, 2026, in Nihonbashi, Tokyo. Details of the event will be announced again shortly.

Representative's Comment

Hiroki Nakahara, Representative Director, President, CEO & CTO of Tokyo Artisan Intelligence

With the completion of the evaluation of the test chip "Sting Ray" based on our proprietary architecture, our cutting-edge technology has taken a major step towards social implementation. The foundation for the next stage, mass production, and global expansion is already being laid. Through this project, we will lead the world's GX and build a cutting-edge AI semiconductor platform that will change the world from Japan.

Company Overview

* Company Name: Tokyo Artisan Intelligence Co., Ltd. * Representative Director, President, CEO & CTO: Hiroki Nakahara * Established: March 3, 2020 * Location: 14th Floor, Shin-Yokohama Square Building, 2-3-12 Shin-Yokohama, Kohoku-ku, Yokohama, Kanagawa Prefecture * Business Activities: Research and development of deep learning algorithms, development and sales of edge AI products, training of AI expert engineers * Company HP: https://tokyo-ai.co.jp/