Name: Lumai Iris
Brand: Lumai

50x

Increase in AI Performance

Lumai’s processor is many times faster than transistor-only accelerators. With a roadmap to deliver 50x the performance, it is the future of AI.

10%

of Total Cost of Ownership

Using 3D optical components means we don't rely on expensive, power-hungry silicon only technology. This results in much lower costs –both capital and operational.

90%

Reduction in Power

Not only is our accelerator fast, it is extremely energy efficient, offering a TOPS/W efficiency that far exceeds existing solutions.

Datacenter-Scale Optical AI Inference

10-100x leap in performance and efficiency

Ultra-efficient with no liquid cooling required

Supports billion-parameter LLMs in real time

Demonstrated with Llama today

Extreme data throughput

Eliminates throughput constraints at scale

Compute in 3D volume not 2D chip

Massive spatial parallelism

Production ready at scale

Built on industry-standard processes - no new materials required

Designed for long-term scaling

Designed for hyperscale workloads

Efficient by design

Parallelism enables millions of operations per cycle

Memory optimized for AI

Three-tier architecture enables efficient, high-capacity KV Cache

Hyperscale ready

Seamless integration into data center environments

Scalability

Extremely wide interface provides high data throughput and amortizes data conversion overhead.

Maturity

Three-tier memory architecture enables efficient & high-capacity memory access for
KV Cache

Performance

Hardware-aware quantization schemes with optical constraints for robust optical inference

Lumai Iris

50x

10%

90%