Nibub Research Initiative

Project Lumina

Pioneering the future of processing through optical computing. Harnessing the power of photons to achieve light-speed computation with unprecedented energy efficiency.

Speed of Light in a Vacuum
299,792,458
meters per second (c)
Energy Consumption
~90%
reduction vs traditional silicon
Latency
O(1)
constant time transmission
Bandwidth Capacity
100+
Terabits per second per waveguide

How Optical Computing Works

Traditional computers push electrons through copper wires. Optical computers guide photons (light particles) through microscopic silicon waveguides.

1

Generation

A tiny, highly efficient laser generates continuous light. This acts as the "power source" for the computation, replacing traditional electrical voltage.

2

Modulation

Data (1s and 0s) is encoded onto the light beam using an optical modulator, which slightly changes the properties of the light (like its phase or amplitude) to represent information.

3

Computation (Interference)

The guided light beams intersect within a specialized matrix. Because photons don't naturally interact with each other, they pass through or interfere in specific, calculable ways. This physical interference mathematically performs matrix multiplication instantly.

4

Detection

Photodetectors at the end of the circuit read the resulting light intensity, converting the answer back into a digital electronic signal that standard computer components can use.

Electrons vs. Photons

Electrons (Traditional) High resistance, generates heat
Photons (Optical) Zero resistance, parallel wavelengths

Optical Physics Sandbox

Interact with a real-time ray-tracing physics simulation. Direct lasers through lenses, bounce them off mirrors, and divide them with splitters to understand how light is manipulated for computation.

Lab Controls
• Pan the camera by dragging empty space.
• Open the Folder icon for Preset Circuits.
• Use the top toolbar to add new components.
• Save/Load JSON configurations with top icons.

Matrix Multiplication Simulation

Run a simulated large-scale neural network matrix multiplication to compare the performance and energy usage of a standard Silicon GPU vs our Lumina Photonic Core.

Task: Process 10B Parameters

Lumina Photonic Core (Light)

Time: 0s Energy: 0 pJ

Standard Silicon GPU (Electrons)

Time: 0s Energy: 0.0 pJ

The Three Pillars of Photonic Computing

Uncapped Velocity

Photons travel at 299,792 km/s. By processing data optically, we eliminate the latency inherent to electronic resistance and capacitive limits, achieving the ultimate theoretical speed limit defined by physics.

Wavelength Multiplexing

Optical paths can cross without interference. By utilizing different colors (wavelengths) of light, multiple immense data streams can be processed simultaneously within the exact same physical space without crosstalk.

Thermal Independence

Electrons moving through copper generate immense heat. Photons moving through waveguides do not. This drastically reduces the energy required to compute, and entirely eliminates the need for massive data center cooling infrastructure.

LUMINA-ARCH-V1 Research Stage
Photonic Matrix Multiplication O(1) latency
Mach-Zehnder Interferometers 100+ GHz
Waveguide Interconnects Terabit/s
Thermal Dissipation -90% vs Si

Transitioning from Silicon to Silica

The semiconductor industry is approaching the fundamental atomic limits of Moore's Law. Transistors are as small as they can physically get without quantum tunneling breaking them.

At Nibub, Project Lumina is accelerating the shift from electrons to photons. By utilizing advanced silicon photonics, we are integrating optical components directly onto traditional chips. Our research focuses heavily on optical neural networks capable of executing heavy machine learning inference (like LLMs) at the speed of light.

Partner with our research team