A System-Level Leap in AI Silicon
OptoML’s innovation lies not in one breakthrough—but in how multiple innovations converge to form a unified, production-ready AI platform.
1. Analog + Optical Fusion
We’re the first to combine analog compute-in-memory with optical interconnect, enabling AI inference with up to 50x better energy efficiency.
2. Fully Integrated AI SoC
While other players offer co-processors, we deliver an entire edge inference engine: NPU, CPU, optional GPU, memory interface, and I/O—on a single chip.
3. Open, Modular Design
RISC-V support, customizable core configurations, and SDK-ready toolchains make us developer-friendly and license-light.
4. Built for the Edge
From smart devices to drones, our chips run AI in real time—without relying on the cloud or cooling systems.
5. Proven and Scalable
We’ve already demonstrated a working analog prototype on 130nm. Our FinFET test chip will validate the full stack in Q2 2025.
OptoML is not just a chip. It’s the new baseline for sustainable, embedded intelligence.
AI is becoming central to everyday devices, from smart home assistants to industrial IoT systems. However, the compute required for deep neural network (DNN) inferencing is highly energy-intensive, contributing to rising emissions and power consumption across billions of edge devices. Today, inferencing makes up over 80% of AI’s total energy use, and global AI workloads are projected to consume more than 29.3 TWh annually—equivalent to Ireland’s electricity consumption.
This creates a major barrier to deploying AI in low-power, resource-constrained environments. The problem is particularly critical at the edge—where real-time intelligence is most valuable, but power budgets are limited.
Data movement between memory and compute is a major source of inefficiency, consuming up to 100x more energy than the compute itself (Mark Horowitz, ISSCC 2014). Conventional AI accelerators also suffer from low compute occupancy, further compounding the energy waste.
Without fundamentally rethinking AI chip design, scaling AI across billions of devices will remain environmentally unsustainable.
optoML is building an ultra-efficient AI System-on-Chip (SoC) that uses patented analog in-memory compute to perform inference directly within memory arrays—dramatically reducing energy and latency.
The SoC integrates our analog NPU alongside SuperScalar CPUs and Vector-GPU cores, offering a complete on-device AI compute solution in a compact and cost-effective form factor. We intent to support software that enables support for open hardware ecosystems.
optoML unlocks real-time, energy-efficient AI at the edge—without reliance on the cloud.