Can Great Sky’s Light-Connected AI Enable Bigger Brains?

2026-07-22 · Source: Big Data & AI News - EE Times · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Cloud Computing & IT Infrastructure · Depth: Expert, extended

Summary

Great Sky, a Boulder, Colorado-based company, is developing AI servers that integrate superconducting neural hardware with photonic interconnects for neuromorphic computing. CEO Dr. Jeff Shainline details systems utilizing Josephson junctions and SQUIDs as core computational elements, offering extremely fast (300 GHz) and energy-efficient processing, alongside robust co-located memory for synaptic weights. Photonic interconnects, employing semiconductor diodes and superconducting nanowire single-photon detectors (SPDs), facilitate low-latency, single-photon-level communication across large scales. Manufacturing leverages conventional lithography with superconducting materials like niobium, targeting 65-45 nanometer nodes for cost-effectiveness. Operating at 4 Kelvin with established helium-4 re-liquefier systems, Great Sky projects brain-scale networks (100 trillion synapses) consuming approximately one megawatt, achieving speeds a million times faster than a human brain. This approach promises a 300x lower cost of ownership for large language models like Llama 405B and enables processing 60 million video frames per second for data-intensive tasks. The company has demonstrated individual components and a 1,000-parameter chip, with a roadmap for multi-chip and wafer-scale integration.

Key takeaway

For AI Architects planning future large-scale AI infrastructure, Great Sky's superconducting optoelectronic platform presents a compelling alternative to semiconductor-based systems. You should evaluate its projected 300x lower cost of ownership for LLMs and its ability to process 60 million video frames per second. While requiring 4 Kelvin cryogenic operation, its flat power scaling curve for massive systems fundamentally alters the economics and capabilities for brain-scale AI, enabling previously unfeasible applications.

Key insights

Superconducting optoelectronic neuromorphic hardware offers extreme speed, energy efficiency, and scalability for brain-scale AI by overcoming communication bottlenecks.

Principles

• Josephson junctions offer highest speed/energy quotient.
• Co-locate memory and processing with superconductors.
• Largest cognitive systems communicate at light speed.

Method

Build neural circuits using SQUIDs for adaptive transfer functions. Integrate superconducting electronics with photonics via semiconductor diodes for light generation and SPDs for single-photon detection.

In practice

• Implement Llama 405B for 300x lower cost of ownership.
• Process 60 million video frames/second for data-intensive AI.
• Develop universal assistants for smart glasses.

Topics

• Superconducting AI
• Neuromorphic Computing
• Photonic Interconnects
• Josephson Junctions
• Single-Photon Detectors
• Cryogenic Computing

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Architect

Related on AIssential

• 😺 🎙️ Watch: Is Brain-like Computing What's Next? — Brain-inspired hardware using light and integrated memory-compute can overcome traditional AI bottlenecks.
• Coherent Breaks Ground on Expanded Texas Facility, Scaling AI’s Optical Backbone — Optical interconnects, leveraging 6-inch indium phosphide wafers, are fundamental for scaling high-performance AI infrastructure efficiently.
• Brain-inspired chip runs near absolute zero and could transform quantum computing — A brain-inspired SiC chip operates near absolute zero, enabling integrated quantum computing control and deep space applications.
• This Breakthrough Could Kill the GPU Race — Optical computing with active metasurfaces offers a path to dramatically reduce AI's energy footprint by performing computation directly with light.
• Will Photonics Loosen the Bottlenecks in Neural Systems? — Optical neural networks using SOAs offer ultra-low-latency, all-optical computation with inherent noise compression and signal regeneration.
• How SpaceXAI & Anthropic Compute Deal Raises Claude Limits — Access to massive compute infrastructure is critical for scaling AI model development and meeting surging user demand.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Big Data & AI News - EE Times.