ULOG3 develops neuromorphic chips that enable onboard AI processing for satellites, supporting both convolutional neural network (CNN) and spiking neural network (SNN) inference.
Funding
Funding not disclosed
Founders
Product
Problem
Satellites have limited power and bandwidth, making it difficult to run conventional AI models onboard for real-time Earth observation and communications tasks. Memory bottlenecks and synchronous processing in traditional AI accelerators lead to high energy consumption, which satellites cannot sustain.
Solution
ULOG3 is developing a neuromorphic processor that can execute both convolutional neural networks (CNN) and spiking neural networks (SNN) directly on satellite hardware. The chip is designed for ultra‑low power consumption, enabling real‑time inference without the need to downlink large data sets. By leveraging event‑driven SNN architectures, the processor reduces memory usage and eliminates the synchronous processing overhead of standard AI accelerators. The first‑generation prototype, created in partnership with the Chennai Institute of Technology, targets a tape‑out before 2026 and will be validated on FPGA and in stratospheric tests. This approach aims to provide satellites with on‑board AI capabilities while preserving their limited power budget.
Target Audience
Primary customers are satellite manufacturers, Earth‑observation mission operators, and communications satellite providers seeking low‑power, on‑board AI processing capabilities.
Features
- Dual inference support for both CNN and spiking neural network models
- Event‑driven neuromorphic architecture that minimizes memory bandwidth and power draw
- Custom ASIC design optimized for space‑qualified radiation tolerance and thermal constraints
- Prototype implementation on FPGA for rapid testing and validation in stratospheric environments
- Planned tape‑out of the first production chip before 2026, with scalability to increase neuron count