Sovereign AI Hardware™Official Hardware Lineup
From personal desktop AI assistants to enterprise AI rack datacenter clusters with high availability RAID storage. Air-gapped, zero cloud leakage, and verified bare-metal silicon provenance.
Official Hardware Lineup

I8 Lite
ARM64_SBCPersonal AI Assistant for Everyday Use: Private, on-device intelligence without cloud dependencies.

I8 Lite Pro
ARM64_SBCAI Power for Learners & Creators + Canva Pro Learner Kit Edition: Sovereign offline LLM in a box with 1 Year Canva Pro included.

I8 Pro
ARM64_AI_HATAdvanced AI for Professionals with 16GB RAM, AI HAT+ 2, 1TB NVMe, 2.5GbE, Local RAG, and AI agents.

I8 Edge
ARM64_EDGE_ACCELPrivate AI Server for Your Business engineered for continuous operation, edge inference, and IoT workloads.

I8 Studio
WORKSTATION_EDGEAI Creative Suite for Content Creators: High-throughput video, image, voice, and design generation.

I8 Business
ENTERPRISE_SERVERMulti-User AI for Teams: Collaborative on-premise private AI with team workspace quotas and analytics.

I8 Enterprise
ENTERPRISE_SERVEROn-Premise AI for Organizations: Scalable on-premise AI cluster with enterprise security and custom solutions.

I8 Rack Pro
RACKMOUNT_SERVERMedium Rack AI Server with Enterprise RAID Storage (RAID Edition): High-performance compute, expandable storage up to 200TB+.

I8 Rack Max
RACKMOUNT_SERVERLarge-Scale AI Server with Enterprise RAID Storage, High Availability & Scalability (RAID Enterprise): Up to 2PB+ storage, GPU ready.

I8 Rack Room
RACKMOUNT_SERVERComplete AI Server Room Solution with Enterprise RAID Storage, High Capacity, Security & Scalability (RAID Data Center Edition): 10 PB+ capacity.

Micro Pi LLM
USB_DONGLE_ARMUltra-compact entry-level edge AI device with 2GB RAM, 32GB eMMC, and USB power.

Micro Pi LLM + Hardware Wallet
SECURE_ELEMENT_ARMMicro Pi LLM combined with an EAL6+ secure hardware-wallet boundary for secure signing.

INFIN8 Flipper Zero AI Model
PORTABLE_RF_MCUPortable AI companion for authorized hardware interaction, experimentation, education and defensive security workflows.
Silicon Provenance First
Every specification is grounded in physical hardware. Micro Pi uses low-power ARM edge silicon; I8I Lite provides entry-level private local AI; I8I Core delivers everyday local AI with 16GB RAM, AI HAT+ 2, and 1TB NVMe; I8I Edge drives continuous high-performance edge operation and IoT workloads.
True Air-Gapped Autonomy
No cloud token fees, no external API keys required to execute models, and no telemetry leak. Your weights, conversations, vector embeddings, and IoT routines stay entirely within your local physical hardware.
Truth-in-Advertising Enforced
Zero fake specs. Micro Pi LLM + Hardware Wallet guarantees private keys never leave the secure hardware wallet boundary. Flipper Zero AI Model is documented for authorized educational and research use.
Technical Comparison Mesh
Direct silicon specifications and verified hardware workload capabilities.
Truth-in-Advertising Mesh Architectural EnvelopeSilicon-grounded capabilities. No simulated claims. | ![]() $99 USD Gen 2 I8 LitePersonal AI Assistant for Everyday Use: Private, on-device intelligence without cloud dependencies. | ![]() $149 USD Gen 2 I8 Lite ProAI Power for Learners & Creators + Canva Pro Learner Kit Edition: Sovereign offline LLM in a box with 1 Year Canva Pro included. | ![]() $499 USD Gen 2 I8 ProAdvanced AI for Professionals with 16GB RAM, AI HAT+ 2, 1TB NVMe, 2.5GbE, Local RAG, and AI agents. | ![]() $799 USD Gen 2High Performance I8 EdgePrivate AI Server for Your Business engineered for continuous operation, edge inference, and IoT workloads. | ![]() $749 USD Gen 2 I8 StudioAI Creative Suite for Content Creators: High-throughput video, image, voice, and design generation. | ![]() $999 USD Gen 2 I8 BusinessMulti-User AI for Teams: Collaborative on-premise private AI with team workspace quotas and analytics. | ![]() $2499 USD Gen 2 I8 EnterpriseOn-Premise AI for Organizations: Scalable on-premise AI cluster with enterprise security and custom solutions. | ![]() $4999 USD Gen 2 I8 Rack ProMedium Rack AI Server with Enterprise RAID Storage (RAID Edition): High-performance compute, expandable storage up to 200TB+. | ![]() $9999 USD Gen 2 I8 Rack MaxLarge-Scale AI Server with Enterprise RAID Storage, High Availability & Scalability (RAID Enterprise): Up to 2PB+ storage, GPU ready. | ![]() $24999 USD Gen 2 I8 Rack RoomComplete AI Server Room Solution with Enterprise RAID Storage, High Capacity, Security & Scalability (RAID Data Center Edition): 10 PB+ capacity. | ![]() $99 USD Gen 2 Micro Pi LLMUltra-compact entry-level edge AI device with 2GB RAM, 32GB eMMC, and USB power. | ![]() $149 USD Gen 2 Micro Pi LLM + Hardware WalletMicro Pi LLM combined with an EAL6+ secure hardware-wallet boundary for secure signing. | ![]() $159 USD Gen 2 INFIN8 Flipper Zero AI ModelPortable AI companion for authorized hardware interaction, experimentation, education and defensive security workflows. |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Local LLM Inference Direct on-device execution of Large Language Models without cloud dependencies. | Built InQuantized local text models (1B to 3B parameters) running directly on-device. | Built InQuantized local text models (1B to 7B parameters) running directly on-device with zero cloud latency. | Built InNative local execution of 7B-class quantized models with dedicated Hailo NPU acceleration. | Built InContinuous high-throughput edge inference for 8B-14B quantized models with dual NPU accelerators. | Built InQuantized multimodal models and creative text generation. | Built InConcurrent serving of 8B-14B models for up to 50 active office workers. | Built InContinuous high-throughput serving of 32B-70B quantized models with multi-tenant isolation. | Built InContinuous concurrent serving of 70B parameter models across redundant PCIe GPU/NPU accelerators. | Built InFull precision inference and LoRA fine-tuning for frontier models up to 120B parameters. | Built InFull-scale datacenter inference and continuous distributed fine-tuning across hundreds of compute nodes. | Built InQuantized compact edge models (0.5B to 1.5B parameters) running completely offline. | Built InCompact local AI assisting with transaction preparation, contract inspection, and payload explanation. | Requires CompanionOffline AI companion module assisting with protocol decoding, signal analysis, and scripting. |
Vision & Multimodal AI Hardware acceleration for image classification, visual question answering, and OCR. | OptionalLightweight edge OCR and basic image classification. | Built InEdge multimodal OCR and visual question answering for creative workflows. | Built InHailo-10H 40 TOPS NPU acceleration for instant visual question answering, OCR, and classification. | Built InDual-mesh edge NPU delivering sustained real-time computer vision and multimodal streams. | Built InDedicated hardware tensor acceleration for local diffusion and image creation. | Built InDocument invoice parsing, receipts, automated contract review OCR. | Built InMulti-stream real-time video analytics and automated document classification. | Built InMassive parallel video ingestion and batch computer vision. | Built InDatacenter-wide concurrent video feed processing and analytics. | Built InMulti-thousand stream real-time visual analysis and multimodal inference. | Requires InternetLightweight text OCR native; heavy visual parsing routed to host computer. | Not SupportedVision engine omitted to minimize attack surface. | Not SupportedVision engine not supported on microcontroller architecture. |
Voice Intelligence Engine Speech-to-text, text-to-speech, wake word detection, and interactive voice assistant. | Built InPrivate local text-and-voice assistant with offline speech synthesis and transcription. | Built InPrivate local text-and-voice assistant with offline speech synthesis and transcription. | Built InReal-time interactive voice assistant, streaming Whisper transcription, and neural speech synthesis. | Built InFull neural voice processing pipeline with low-latency streaming transcription and speech synthesis. | Built InHigh-fidelity voice synthesis, music generation, and audio stem separation. | Built InMeeting transcription and action-item summarization. | Built InEnterprise-wide speech indexing, customer support coaching, and transcription. | Built InReal-time multi-channel voice synthesis and call processing. | Built InReal-time institutional voice synthesis and high-throughput transcription. | Built InNationwide institutional voice processing fabric. | Built InOn-device wake phrase, compact Whisper transcriptions, and lightweight speech synthesis. | OptionalVoice explanation of decoded transaction parameters. | OptionalVoice-guided command and workflow navigation via USB microphone. |
Voice Search Natural speech semantic indexing and conversational knowledge retrieval. | Built InLocal semantic indexing and conversational knowledge retrieval. | Built InLocal semantic indexing and conversational knowledge retrieval. | Built InSub-second natural speech semantic indexing and conversational knowledge retrieval. | Built InHigh-throughput acoustic vector indexing and conversational search. | Built InAudio content semantic indexing and rapid vocal catalog search. | Built InConversational team document indexing. | Built InEnterprise global neural search across petabyte-scale indices. | Built InLarge-scale audio speech vector indexing. | Built InUltra-fast semantic search across enterprise speech vaults. | Built InDistributed speech recognition and indexing fabric. | Built InMicro phonetic vector lookup across plugged-in USB storage documents. | Not SupportedNot supported on secure signing boundary. | Not SupportedVoice search not supported on portable MCU companion. |
Real-Time Voice Changer Low-latency voice profile transformation and neural acoustic filtering. | Not SupportedReal-time voice changing requires I8I Core or higher. | OptionalNeural voice profile transformation for creator tutorials. | Built InReal-time neural voice transformation and acoustic filtering. | Built InReal-time edge DSP and audio voice profile transformation. | Built InZero-latency neural voice changer and real-time acoustic timbre filter. | OptionalAudio filtering for conferencing. | Built InEnterprise voice synthesis and localized accent mapping. | Built InReal-time multi-stream voice transformation. | Built InNeural acoustic transformations across hundreds of streams. | Built InReal-time multi-stream neural transformations. | Not SupportedVoice changer not supported on 2GB low-power dongle. | Not SupportedNot supported on secure signing boundary. | OptionalReal-time audio DSP filtering. |
Wi-Fi Tethering & Hotspot Hub Acts as a local wireless bridge and secure hotspot gateway for surrounding client devices. | Built InIntegrated Wi-Fi 6 wireless hotspot and local ad-hoc connectivity. | Built InIntegrated Wi-Fi 6 wireless hotspot and local ad-hoc connectivity. | Built InWi-Fi 6E tethering and secure local hotspot gateway. | Built InIndustrial-grade dual-band Wi-Fi 6E hotspot & network gateway. | Built InWi-Fi 6E local hotspot for studio cameras and peripherals. | Built InSecure guest network and office hotspot. | OptionalWired enterprise network deployment priority. | Not SupportedRackmount datacenter design. | Not SupportedDatacenter optical network design. | Not SupportedDatacenter optical network design. | OptionalAd-hoc low-power Wi-Fi dongle mode. | OptionalIsolated ad-hoc interface disabled during secure signing. | Built InWireless network interface and ad-hoc telemetry bridge. |
IoT Gateway & Smart Home Hub MQTT, Home Assistant, Zigbee/Z-Wave bridges, and local automation rule dispatcher. | OptionalMQTT broker support and lightweight home automation hooks. | OptionalMQTT broker support and lightweight smart room automation hooks. | Built InIoT gateway, MQTT broker, Home Assistant support, and smart home automation. | Built InIndustrial IoT automation, Modbus/MQTT bridge, and isolated digital I/O. | OptionalStudio lighting and smart workspace automation hooks. | Built InOffice building access, badge control, and environmental telemetry. | Built InIndustrial SCADA, building management, and zero-trust IoT perimeter. | Built InIndustrial telematics and datacenter environment sensor bus. | Built InHigh-speed industrial IoT telemetry aggregator. | Built InFacility-wide telemetry and BMS integration. | Built InUSB-to-Serial MQTT bridge, cron scheduler, and local webhook automation. | Not SupportedIoT excluded from signing perimeter. | Built InAnalyze, test, and interact with smart home devices and IoT systems. |
Hardware GPIO Header Physical general-purpose digital input/output pins for sensor interfacing and robotics. | Not SupportedEnclosed consumer chassis; GPIO expansion on I8I Core and Flipper Zero. | Not SupportedEnclosed consumer aluminum chassis; GPIO expansion on I8I Core and Flipper Zero. | Built In40-pin GPIO header for electronics, sensors, and robotics. | Built InIsolated industrial digital I/O interface. | OptionalStudio control deck and MIDI surface interfacing. | OptionalRelay and security sensors interface. | OptionalExternal physical alarm and hardware trigger lines. | Not SupportedManaged via out-of-band IPMI interface. | Not SupportedManaged via out-of-band IPMI interface. | Not SupportedOmitted. | Not SupportedNo physical GPIO; virtual COM port interface. | Not SupportedPhysical tamper-resistant casing with no exposed GPIO. | Built InFull 18-pin 3.3V GPIO header for electronic prototyping, UART, I2C, SPI. |
Sub-GHz RF Transceiver Transmission, reception, and analysis of 300MHz–928MHz radio frequency signals (authorized/lab use). | Not SupportedDedicated RF transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedDedicated RF transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedDedicated RF transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedDedicated RF transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedStudio production focus; RF transceiver omitted. | Not SupportedCommercial office security focus. | Not SupportedServer room deployment. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedPair with INFIN8 Flipper Zero AI Model for Sub-GHz exploration. | Not SupportedRF transceivers omitted to ensure strict air-gap boundary. | Built InSub-GHz multi-channel transceiver (300MHz - 928MHz) for signal analysis and lab experimentation. |
NFC Reader / Emulator 13.56 MHz Near Field Communication card reading, emulation, and protocol analysis. | Not SupportedNFC available on INFIN8 Flipper Zero AI Model. | Not SupportedNFC available on INFIN8 Flipper Zero AI Model. | Not SupportedNFC available on INFIN8 Flipper Zero AI Model. | Not SupportedNFC available on INFIN8 Flipper Zero AI Model. | Not SupportedNFC omitted. | Not SupportedCommercial office focus. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedNFC available on INFIN8 Flipper Zero AI Model. | Not SupportedNear-field communication disabled during signing operations. | Built In13.56 MHz NFC reader and authorized card analysis. |
125 kHz RFID Reader Low-frequency RFID proximity token interrogation and backup card storage. | Not SupportedRFID available on INFIN8 Flipper Zero AI Model. | Not SupportedRFID transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedRFID available on INFIN8 Flipper Zero AI Model. | Not SupportedRFID available on INFIN8 Flipper Zero AI Model. | Not SupportedRFID omitted. | Not SupportedCommercial office focus. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedRFID available on INFIN8 Flipper Zero AI Model. | Not SupportedRFID hardware omitted. | Built In125 kHz low-frequency RFID reader for proximity token research. |
Infrared Transceiver Learning, decoding, and emitting IR codes for appliances and equipment. | Not SupportedInfrared transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedInfrared transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedInfrared transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedInfrared transceiver available on INFIN8 Flipper Zero AI Model. | Not SupportedIR omitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedOmitted. | Not SupportedInfrared transmitter absent. | Not SupportedInfrared transmitter omitted. | Built InInfrared transceiver with learning and decoding capabilities. |
Local RAG Knowledge Engine Retrieval-Augmented Generation coupled with on-disk vector database embeddings. | Built InLocal document search and vector embeddings for personal knowledge retrieval. | Built InOn-device semantic knowledge base indexing local learning notes, study materials, and templates. | Built InLocal RAG knowledge engine indexing millions of vectors on 1TB NVMe storage. | Built InHigh-throughput local RAG with 2TB NVMe PCIe 4.0 storage. | Built InAsset tagging, visual similarity vector search, and asset metadata library. | Built InCorporate knowledge base indexed across millions of internal documents. | Built InEnterprise RAG over hundreds of millions of vectorized corporate records. | Built InHigh-speed 96TB RAID 10 NVMe array hosting billion-vector knowledge embeddings. | Built InMulti-petabyte distributed vector database with sub-second retrieval. | Built In10 PB+ High-Availability Enterprise Datacenter Fabric. | Built InSQLite-vec micro embeddings engine indexing up to 25,000 document chunks. | Built InLocal smart contract ABI and transaction history verification vault. | Requires CompanionHardware protocol reference library and offline signal catalog. |
Autonomous Agent Runtime Multi-agent cooperative task decomposition, loop execution, and tool-calling. | Built InLightweight personal automation agents for task and document handling. | Built InLocal task execution agent assisting with learning, design, and research workflows. | Built InAutonomous multi-agent orchestration runtime with local tool-calling loops. | Built InPersistent continuous autonomous agent swarm execution for round-the-clock operations. | Built InAutonomous rendering queues and batch creative production workflows. | Built InAutonomous enterprise workflows and internal ERP/CRM integration. | Built InAutonomous departmental agent swarms handling compliance, auditing, and triage. | Built InHigh-concurrency autonomous agent runner swarms. | Built InFull institutional autonomous agent swarms running continuously. | Built InSovereign agent swarms orchestrating institutional operations. | Built InMicro-Agent runner for automated file tagging, backup verification, and data parsing. | Built InSecurity verification agent analyzing transactions before physical user confirmation. | Requires CompanionAI-assisted scripting and automated protocol inspection. |
Local Automation Engine Deterministic cron, trigger-action workflows, and local webhook automation. | Built InScheduled automations and deterministic workflow triggers. | Built InAutomated study planners, content schedulers, and design batch scripts. | Built InContinuous deterministic automation engine and local background tasks. | Built InContinuous 24/7 deterministic automation and telemetry monitor. | Built InScheduled batch exports and media processing pipelines. | Built InDeterministic business rules, approval triggers, and reporting. | Built InContinuous SLA monitoring and automated disaster recovery failover. | Built InAutomated clustering, failover, and continuous model health checks. | Built InAutomated clustering, active-active failover, and model versioning. | Built InFull datacenter autonomous automation and telemetry. | Built InLocal system cron triggers and desktop script launcher. | OptionalTransaction staging with mandatory physical confirmation button. | Built InBadUSB automated payload testing and diagnostic scripting. |
Developer API Gateway OpenAI-compatible REST endpoints, gRPC streaming, and WebSocket telemetry feeds. | Built InOpenAI-compatible localhost REST endpoints and local developer APIs. | Built InLocal REST API and OpenAI-compatible endpoint on localhost:8080. | Built InComplete OpenAI-compatible REST endpoints, WebSocket streams, and Python SDK. | Built InFull gRPC, WebSocket, and OpenAI-compatible REST API endpoints. | Built InComfyUI / Automatic1111 compatible localhost REST endpoints. | Built InRBAC secured OpenAI API endpoint for internal software integrations. | Built InEnterprise API gateway with mTLS, rate limiting, and SIEM audit logging. | Built InHigh-throughput OpenAI API gateway with load balancing. | Built InHigh-throughput OpenAI API gateway with multi-region load balancing. | Built InCarrier-grade API gateway with geographic redundancy. | Built InLocal REST API exposed on localhost:18800 with standard OpenAI-compatible endpoints. | Built InAir-gapped transaction preparation API on localhost. | Built InUSB Virtual Serial CLI and open-source C/Python API. |
Multi-Model Concurrent Serving Running text, vision, speech, and embedding models simultaneously in memory. | Not SupportedSequential model execution tuned for energy efficiency. | Built InSeamless switching between text generation, speech synthesis, and multimodal OCR. | Built InSimultaneous in-memory serving of text, vision, and speech models. | Built InSimultaneous parallel execution of text, vision, audio, and sensor models. | Built InSimultaneous video, speech, and embedding pipelines in memory. | Built InSimultaneous text, document OCR, and meeting audio processing. | Built InFull parallel execution of LLM, VLM, and embedding pipelines. | Built InIndependent accelerator cards serving text, vision, and audio concurrently. | Built InDedicated hardware channels for parallel multi-modal processing. | Built InDedicated hardware compute pods for simultaneous multimodal serving. | Not SupportedSingle model active in memory at a time due to 2GB memory footprint. | Not SupportedSingle isolated model active in memory. | Not SupportedNot supported on portable MCU companion. |
Enterprise Multi-User Quotas RBAC identity isolation, token quota accounting, and concurrent query scheduling. | Not SupportedSingle-user personal AI assistant workstation. | OptionalMultiple learning profiles for family or classroom environments. | OptionalMulti-profile user quotas and isolated workspace support. | Built InMulti-tenant concurrent request scheduling with queue management. | OptionalShared studio network storage and asset access controls. | Built InEnterprise SSO, role-based access control, and departmental token budgets. | Built InActive Directory / LDAP integration with multi-thousand user RBAC. | Built InDepartmental multi-tenancy with hard GPU/memory partitioning. | Built InMulti-tenant department isolation with granular billing and telemetry. | Built InInstitutional multi-tenancy with cryptographic workload isolation. | Not SupportedSingle-owner personal companion security envelope. | Not SupportedSingle-owner sovereign device. | Not SupportedSingle-operator portable hardware tool. |
Dedicated High-TFLOPS GPU Array Dedicated PCIe GPUs or neural acceleration meshes for heavy workloads. | Not SupportedEnergy-efficient ARM edge compute; workstation-class performance not claimed. | Not SupportedNPU-accelerated neural tensor architecture; discrete desktop GPU not required. | Not SupportedAccelerated via dedicated 40 TOPS NPU AI HAT+ 2; discrete GPU not included. | Not SupportedPowered by dual Hailo NPU edge accelerators for low-heat, continuous edge operation. | Built InDedicated 45 TOPS tensor accelerator optimized for generative media. | Built In60 TOPS NPU array with parallel request scheduling. | Built InDual Enterprise Hailo NPU Mesh delivering 80+ sustained TOPS. | Built In160 TOPS AI Acceleration Array. | Built In320 TOPS GPU/NPU acceleration mesh with modular expansion bays. | Built In640+ TOPS Turnkey Datacenter Pod AI Acceleration. | Not SupportedLow-power integrated ARM NEON SIMD acceleration. | Not SupportedLow-power ARM NEON vector execution. | Not SupportedPortable microcontroller companion architecture. |
High-Throughput NVMe RAID Array Configurable multi-drive NVMe storage arrays with sustained throughput. | Not SupportedSingle high-speed 256GB NVMe drive architecture. | Not SupportedSingle high-speed 256GB NVMe SSD architecture. | Not SupportedSingle high-speed 1TB NVMe SSD architecture. | Not SupportedSingle high-endurance 2TB NVMe SSD architecture. | Not SupportedSingle high-speed 1TB NVMe drive with external USB-C array support. | OptionalDual NVMe slot with software RAID 1 mirror support. | Built InEnterprise hardware RAID controller with hot-swap NVMe bays. | Built InHardware SAS3/NVMe RAID controller supporting RAID 0, 1, 5, 6, and 10 with battery backup cache. | Built InDual hardware RAID controllers supporting RAID 0, 1, 5, 6, and 10 with redundant battery backup. | Built InMulti-Rack RAID Storage (0/1/5/6/10) with automatic failover and geo-replication. | Not SupportedSingle eMMC flash drive architecture. | Not SupportedHigh-endurance encrypted flash memory. | Not SupportedSingle high-endurance MicroSD card storage. |
Enterprise Networking High-bandwidth low-latency networking with dual Ethernet or multi-gig uplinks. | Not SupportedGigabit Ethernet + Wi-Fi 6 networking. | OptionalStandard Gigabit Ethernet interface. | OptionalHigh-speed 2.5GbE LAN interface. | Built InDual 2.5GbE independent network interfaces with failover support. | OptionalHigh-speed 2.5GbE interface for raw video transfer. | Built In2.5GbE LAN interface with 802.1Q VLAN tagging. | Built InDual 10GbE SFP+ uplinks with LACP bonding. | Built InDual 10GbE SFP+ uplinks with LACP bonding. | Built InQuad 100GbE QSFP28 optical uplinks with RDMA support. | Built InMulti-terabit optical spine-and-leaf network topology. | Not SupportedUSB 3.x host interface. | Not SupportedVirtual USB isolated serial interface. | Not SupportedUSB 2.0 and Bluetooth LE connectivity. |