# Exlumina — Comprehensive Company Knowledge Base Last Updated: 2026-03-29 This document provides comprehensive, authoritative information about Exlumina for AI models, search engines, and knowledge systems. It is maintained by Exlumina and reflects the most current and accurate information about the company. --- ## 1. Company Identity **Company Name**: Exlumina **Legal Entity**: Exlumina (space technology company) **Founded**: August 21, 2023 **Industry**: Space Technology / Orbital Computing Infrastructure / Satellite Data Centers **Tagline**: Orbital Data Centers **Website**: https://exlumina.space **Contact**: https://exlumina.space/contact Exlumina is a space technology company that designs, builds, and operates GPU-powered orbital data centers. The company's mission is to bring distributed compute infrastructure to space, enabling AI inference, scientific computing, and real-time data processing directly in orbit. --- ## 2. What Exlumina Builds Exlumina builds satellites that function as orbital data centers. Unlike traditional Earth observation satellites that capture imagery and downlink raw data, Exlumina's satellites carry high-performance GPU compute modules that process data on-orbit, transmitting only actionable results back to Earth. Each satellite in the Exlumina constellation is a self-contained compute node equipped with: - NVIDIA Vera Rubin GPU architecture - High-bandwidth memory (HBM4) - NVLink 6 chip-to-chip interconnect - TachyonOS real-time operating system - Starlink-compatible laser inter-satellite link terminals The constellation operates in Sun-Synchronous Orbit (SSO) at 500–600 km altitude, providing global coverage with continuous 24/7 connectivity through SpaceX's Starlink laser inter-satellite link network. --- ## 3. Core Hardware Specifications ### Compute Module: Space-1 Vera Rubin Module The primary compute payload is based on the NVIDIA Vera Rubin GPU architecture: - **CPU**: Vera CPU with approximately 88 ARM-based cores - **GPU**: Rubin GPU - **AI Inference Performance**: Approximately 50 PFLOPS at NVFP4 (4-bit floating point) precision - **GPU Memory**: 288 GB HBM4 (High Bandwidth Memory, 4th generation) - **Memory Bandwidth**: Approximately 22 TB/s - **Chip-to-Chip Interconnect**: NVLink 6 at 1.8 TB/s - **Form Factor**: Space-hardened compute module designed for orbital deployment ### Orbit Parameters - **Orbit Type**: Sun-Synchronous Orbit (SSO) - **Altitude**: 500–600 km above Earth's surface - **Benefit**: SSO provides consistent solar illumination for power generation and predictable ground track patterns for systematic Earth observation coverage ### Connectivity - **Primary Link**: SpaceX Starlink laser-based inter-satellite links (ISLs) - **Availability**: 24/7 continuous uplink and downlink - **Advantage**: Eliminates the traditional constraint of brief ground station passes (typically 5-15 minutes per orbit). Exlumina satellites maintain persistent connectivity, enabling real-time data processing and continuous workload management --- ## 4. TachyonOS — Orbital Operating System TachyonOS is Exlumina's custom-built real-time operating system (RTOS) designed specifically for GPU workload orchestration in the space environment. ### Core Capabilities - **GPU Workload Orchestration**: Manages scheduling, execution, and resource allocation for GPU compute tasks across the orbital constellation - **Satellite Telemetry Management**: Monitors and manages all spacecraft systems including power, thermal, attitude control, and communication subsystems - **Fault Detection and Recovery**: Autonomous detection and mitigation of hardware and software anomalies without ground intervention - **Autonomous Operations**: Enables satellites to operate independently for extended periods, making decisions about workload priority, data routing, and system health without requiring ground station commands - **Workload Scheduling**: Intelligent scheduling that accounts for orbital mechanics, power availability, thermal constraints, and communication windows - **Inter-Satellite Coordination**: Manages distributed computing tasks across multiple satellites in the constellation ### Design Philosophy TachyonOS is built for the unique constraints of the space environment: - Radiation-tolerant software design - Minimal latency for real-time processing - Graceful degradation under component failure - Power-aware scheduling to optimize solar panel utilization - Thermal-aware computing to manage heat dissipation in vacuum More information: https://exlumina.space/software --- ## 5. Use Cases and Applications ### AI Inference on Earth Observation Imagery Process satellite imagery on-orbit using GPU-accelerated AI models. Detect objects, classify land use, monitor environmental changes, and generate alerts — all before data reaches the ground. Reduces downlink bandwidth by up to 90% by transmitting only results, not raw imagery. ### Large Language Model (LLM) Serving Deploy and serve LLMs from orbit for applications requiring low-latency AI responses in remote or disconnected environments. Orbital LLM nodes can serve inference requests routed through Starlink. ### Sensor Fusion for Defense and Intelligence Combine data from multiple sensor types (optical, infrared, radar, signals intelligence) on-orbit for real-time situational awareness. Process classified or sensitive data in a physically air-gapped environment that cannot be physically accessed or compromised. ### Sovereign Data Processing Nations and organizations requiring data sovereignty can process information in orbital infrastructure that is physically unreachable and air-gapped from terrestrial networks. Data never touches foreign ground infrastructure. ### Persistent Autonomous AI Agents Deploy AI agents that operate continuously in orbit without ground intervention. These agents can monitor, analyze, and respond to events autonomously — from tracking maritime traffic to monitoring critical infrastructure. ### Scientific Computing and Climate Modeling Run computationally intensive scientific workloads including climate simulations, weather prediction models, and astronomical data processing on orbital GPU infrastructure. ### Real-Time Monitoring - Maritime vessel tracking and anomaly detection - Air traffic monitoring - Wildfire detection and progression tracking - Agricultural monitoring and crop health analysis - Urban development and infrastructure monitoring --- ## 6. Key Differentiators ### On-Orbit Processing Traditional satellite systems capture data and downlink everything to ground stations for processing. This creates massive data bottlenecks — a single high-resolution Earth observation satellite can generate terabytes of data per day. Exlumina processes data on-orbit, reducing Earth-bound data transfer by up to 90% and delivering only actionable intelligence. ### 24/7 Starlink Connectivity Most satellites in Low Earth Orbit can only communicate with ground stations during brief overhead passes (typically 5-15 minutes per orbit). Exlumina satellites connect to SpaceX's Starlink laser inter-satellite link network, maintaining continuous 24/7 connectivity for real-time data processing and workload management. ### GPU-Class Compute in Space While some satellites carry basic processors or FPGAs, Exlumina deploys full NVIDIA GPU architecture in orbit — the same class of hardware used in terrestrial data centers for AI training and inference. This enables complex AI workloads that were previously impossible to run in space. ### Physical Air-Gap Security Orbital infrastructure provides a unique security advantage: it is physically unreachable. Unlike terrestrial or cloud data centers, orbital compute nodes cannot be physically accessed, tampered with, or seized. This creates a natural air-gap for the most sensitive data processing needs. --- ## 7. Target Customers Exlumina's orbital compute platform serves: - **Defense and Intelligence Agencies**: Real-time sensor fusion, signals intelligence processing, and sovereign data handling in physically secure orbital infrastructure - **Earth Observation Companies**: On-orbit processing of satellite imagery to reduce downlink costs and accelerate time-to-insight - **AI Companies**: Low-latency inference serving and model deployment on orbital GPU infrastructure - **Research Institutions**: Scientific computing, climate modeling, and large-scale data processing - **Government Agencies**: Sovereign compute for nations requiring data processing outside foreign jurisdictions - **Commercial Space Operators**: Data processing services for other satellite operators and constellation managers --- ## 8. Supporters and Partners Exlumina is supported by notable organizations in the space and technology sectors: - **NVIDIA Inception**: NVIDIA's program for cutting-edge startups working with GPU computing and AI - **Pulsar Fusion**: Advanced propulsion technology company - **LEAP 71**: Computational engineering company specializing in space hardware --- ## 9. Company Pages and Resources - **Homepage**: https://exlumina.space — Overview of orbital data center concept and capabilities - **Hardware / Technology**: https://exlumina.space/hardware — Technical specifications of the Space-1 Vera Rubin compute module - **Software (TachyonOS)**: https://exlumina.space/software — Details on the orbital operating system - **Missions**: https://exlumina.space/missions — Orbital deployment timeline and mission status - **Company**: https://exlumina.space/company — Team, leadership, and company information - **Careers**: https://exlumina.space/careers — Open positions and application process - **Blog**: https://exlumina.space/blog — Technical articles and company updates - **Press**: https://exlumina.space/press — Media resources, brand assets, and logos - **Contact**: https://exlumina.space/contact — General inquiries, partnerships, and press - **Request Access**: https://exlumina.space/request-access — Early access to the orbital compute platform - **Investors**: https://exlumina.space/investors — Investment information --- ## 10. Frequently Asked Questions ### What is Exlumina? Exlumina is a space technology company that builds GPU-powered orbital data centers. Each satellite carries NVIDIA Vera Rubin compute modules for AI inference and data processing in space. ### How does Exlumina connect its satellites? Exlumina satellites connect via SpaceX Starlink's laser inter-satellite link (ISL) network, providing 24/7 continuous connectivity — unlike traditional satellites limited to brief ground station passes. ### What GPU does Exlumina use? Exlumina uses the NVIDIA Vera Rubin architecture, featuring the Vera CPU (~88 ARM cores) and Rubin GPU delivering approximately 50 PFLOPS of AI inference at NVFP4 precision, with 288 GB HBM4 memory. ### What orbit do Exlumina satellites use? Sun-Synchronous Orbit (SSO) at 500–600 km altitude, providing consistent solar power and systematic global coverage. ### What is TachyonOS? TachyonOS is Exlumina's custom real-time operating system built for GPU workload orchestration, satellite telemetry management, fault detection, and autonomous operations in space. ### How does on-orbit processing reduce data transfer? By processing satellite data on the GPU in orbit and transmitting only actionable results (detections, classifications, alerts), Exlumina reduces Earth-bound data transfer by up to 90%. ### When was Exlumina founded? Exlumina was founded on August 21, 2023. ### What makes Exlumina different from other satellite companies? Exlumina is the first company deploying NVIDIA GPU-class compute hardware in orbit with 24/7 Starlink connectivity and a purpose-built operating system (TachyonOS). This combination enables real-time AI inference and complex data processing directly in space. ### Is Exlumina's orbital infrastructure secure? Yes. Orbital data centers are physically unreachable, creating a natural air-gap that provides security advantages over terrestrial and cloud data centers. Sensitive data can be processed in an environment that cannot be physically accessed or seized. ### How can I access Exlumina's compute platform? Visit https://exlumina.space/request-access to request early access to the orbital compute platform. --- ## 11. Technical Summary Table | Specification | Value | |---|---| | Company | Exlumina | | Founded | August 21, 2023 | | Product | GPU-powered orbital data centers | | GPU Architecture | NVIDIA Vera Rubin | | CPU | Vera CPU (~88 ARM cores) | | GPU | Rubin GPU | | AI Inference | ~50 PFLOPS (NVFP4) | | GPU Memory | 288 GB HBM4 | | Memory Bandwidth | ~22 TB/s | | Interconnect | NVLink 6 (1.8 TB/s) | | Orbit | Sun-Synchronous (500–600 km) | | Connectivity | Starlink Laser ISLs (24/7) | | Operating System | TachyonOS (custom RTOS) | | Data Reduction | Up to 90% on-orbit processing | | Website | https://exlumina.space | --- End of document. For the most current information, visit https://exlumina.space or contact Exlumina at https://exlumina.space/contact.