Computing at the Speed of Light.

In 2003, a chip's core technology promised 8 trillion operations per second using light. This is the story of the LensLet EnLight256, a concept that was ahead of its time, and whose time has finally come.

8 TOPS

Peak *core* performance in 2003, bottlenecked by the surrounding electronics.

100 POPS

The projected performance of a fully optimized optical architecture like Bluebottle A2.

>25x Efficiency

The fundamental advantage of optical computing over the best modern silicon.

The Core Concept: Vector-Matrix Multiplication with Light

The EnLight256's engine replaced electronic transistors with photons to perform a complex mathematical operation—essential for modern AI—in a single flash of light. Hover over each step to learn more.

💡

Source Array

(Input Vector)

🔳

Spatial Light Modulator

(The Matrix)

📊

Detector Array

(Output Vector)

Hover over a diagram component to see details here.

Performance: A Tale of Three Eras

To logically assess the technology's potential, we make fair comparisons: pitting the components of each era against their state-of-the-art digital counterparts.

Era 1: 2003 - A Glimpse of an Unreachable Future

In 2003, the EnLight256's optical core was astoundingly powerful, but its system-level performance was hamstrung by the slow electronics of the day.

Core Performance (Log Scale)

Core Efficiency (TOPS per Watt)


Era 2: Near Future (2025) - The Promise Realized

With modern interconnects, a contemporary optical processor demonstrates a commanding lead in both raw performance and, most critically, power efficiency.

Projected Performance

Projected Efficiency (TOPS per Watt)


Era 3: The Zenith (2027+) - The 'Bluebottle A2' Vision

The ultimate goal: total integration. By co-designing the optical core with on-chip SRAM (1ns latency) and a 100% compiled machine code executor, all remaining overhead is eliminated, unlocking explosive performance gains.

Projected Performance

Projected Efficiency (TOPS per Watt)

Architectural Evolution: From Bottleneck to Fabric

The journey from the 2003 concept to the Bluebottle A2 vision is a story of systematically eliminating bottlenecks through deeper and deeper integration.

1. THEN (c. 2003 System)

Off-chip DSP & Memory
PCI Bus (~1 GB/s)
Optical Core

The Bottleneck: An ultra-fast core crippled by a slow, external data bus.

2. NOW (Conceptual 2025 System)

On-Chip Controller & Memory
CXL/Optical Bus (>100 GB/s)
Optical Core

The Solution: A high-speed bus allows the core to be properly utilized, unleashing its performance.

3. ZENITH (Bluebottle A2)

Compiled Code Executor

Compute Fabric

(Optical Core + Integrated SRAM)

The Integration: The bus is eliminated. Core and memory merge into a single fabric, driven by compiled code for zero-overhead execution.

From Fixed to Fluid: The Matrix Evolves

Beyond the I/O bottleneck, another key limitation was the difficulty in reprogramming the optical matrix. Modern material science has solved this, enabling true general-purpose optical computing.

THEN (c. 2003)

The original Spatial Light Modulator (SLM) was like a fixed stencil. Changing the matrix for different calculations was slow or impossible, limiting it to specialized tasks.

NOW (The Future)

Modern optical processors use phase-change materials like Vanadium Dioxide (VO₂). These materials can instantly change their optical properties, creating a fully reprogrammable matrix on the fly.

The Trillion-Dollar Opportunity

Optical computing isn't just an improvement; it's a fundamental shift in the economics of AI and high-performance computing.

AI Compute Cost Breakdown (Total Cost of Ownership)

Optical computing drastically reduces energy consumption, which can account for over 50% of the total cost of ownership for large-scale AI infrastructure. This translates to direct, massive savings.

📈

Massive Market

The AI hardware market is projected to exceed $400 billion by 2027. Optical processors are positioned to capture a significant share of this rapidly growing TAM.

🌍

Sustainable AI (ESG)

Dramatically lower power consumption addresses the growing environmental concerns of AI, aligning with global sustainability goals and attracting ESG-focused investment.

🏰

Defensible Moat

This technology represents a paradigm shift, not an incremental improvement. Early movers will build a strong, defensible IP portfolio in a category silicon cannot compete in on efficiency.

The Future is Built with Light

With historical bottlenecks resolved and a clear path to total integration through the Bluebottle A2 vision, the principles of the EnLight256 are ready to redefine computation.

⚡️

Unmatched Speed

Calculations happen at the speed of light, with massive parallelism. This could reduce AI training times from weeks to hours.

❄️

Radical Efficiency

With minimal heat, the energy costs of data centers could plummet, making AI more sustainable and accessible.

💡

Beyond Silicon's Limits

As Moore's Law for electronics slows, optical computing provides a new dimension for performance scaling.