GEMESYS

What GEMESYS does

GEMESYS Technologies develops analog chip designs based on the information-processing mechanisms of the human brain. This novel architecture aims to overcome the inefficiencies of current hardware for artificial intelligence, enabling AI hardware vendors to distribute chips that are significantly more energy-efficient for training neural networks. The technology reduces the cost, time, and data required for AI training, while also improving overall quality and performance. Its small size and high energy efficiency allow for embedded AI capabilities, bringing both AI training and inferencing to edge devices. GEMESYS is also involved in the EMULAITE project, which focuses on developing a revolutionary approach to AI training using analog circuit emulation and memristors to achieve 100–1,000x more energy-efficient AI training, supporting sustainable edge AI and reducing the environmental impact of AI data centers.

Products

• GEMESYS Chip ArchitectureAn integrated circuit that redefines AI-chip architecture by adapting the human brain's information processing mechanisms, enabling high-efficiency AI training and new inference capabilities.
• EMULAITEA revolutionary approach to AI training funded by the Grüne Gründungen.NRW program, utilizing analog circuit emulation and memristors to replace inefficient gradient-based learning, making AI training 100–1,000x more energy-efficient for sustainable edge AI.

Key capabilities

• Fully analog chip design
• Inspired by the human brain's information processing
• 20,000 times more energy-efficient AI training than current technology (expected)
• Enables AI training and inferencing capabilities at the edge
• Significantly reduces cost, time, and data for neural network training
• Increases overall AI quality and performance
• Small chip size
• High data efficiency
• Low power consumption
• Neural networks with high sparsity and low depth for better inference operations
• Memristor-enabled analog paradigm

Use cases

• Edge AI training
• Edge AI inferencing
• Making every device and sensor smart, autonomous, and adaptable
• Sustainable AI
• Reducing environmental impact of AI data centers

AI approach

GEMESYS develops analog chip designs based on the information-processing mechanisms of the human brain to create energy-efficient AI hardware for edge AI training and inference. Their technology aims to significantly reduce the cost, time, and data required to train neural networks, while increasing overall quality and performance. They are developing a novel chip architecture that is fully analog and utilizes memristors to achieve high efficiency.

Tech named: Circuit Synthesis, Digital Emulation, Neuromorphic Computing, Memristive Circuits, analog chip design, neural networks, memristor, integrated circuit, analog circuit emulation

What it says sets it apart

• The only chip on the market capable of bringing both AI training and inferencing capabilities to the edge
• Fully analog and inspired by the human brain
• Overcomes current bottlenecks in information technology with an analog chip design
• Provides the world’s first hardware for edge AI training
• Expected 10,000x increase in training efficiency due to analog paradigm
• Enables AI democratization by making cheap and highly efficient AI chips accessible
• Utilizes memristors for its fully analog design
• EMULAITE project aims for 100-1,000x more energy-efficient AI training through analog circuit emulation and memristors

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.