At COMPUTEX 2025, where edge computing meets real-world AI deployment, Innodisk delivered a showcase not of concepts—but of complete, production-ready ecosystems. Known for its deep specialization in industrial memory, flash storage, and embedded computing, Innodisk brought a compelling narrative this year: AI at the edge is real, local, and scalable—and they’re already shipping the platforms to prove it.
Heterogeneous AI Platforms in Action
Innodisk’s core strength has long been its ability to deliver embedded reliability across fragmented platforms. That strength was on full display this year, with a coordinated demo suite built on solutions from NVIDIA, Intel, and Qualcomm Technologies, seamlessly integrated with Innodisk’s own ecosystem of camera modules, DRAM, SSDs, and expansion modules.
The message was clear: no matter your compute engine, Innodisk delivers plug-and-deploy readiness for AI workloads at the edge.
APEX-X100 and AccelBrain: Private LLM Training, Locally
The showstopper was Innodisk’s enterprise-grade on-premise LLM platform, powered by the APEX-X100 AI computer.
-
Compact yet powerful, the APEX-X100 is built for local AI model training
-
Paired with AccelBrain, Innodisk’s proprietary AI toolchain, it delivers low-latency inference and retraining capabilities without relying on external cloud services
-
Designed for data-sensitive applications across healthcare, manufacturing, and government sectors
For enterprises eyeing sovereign AI strategies—this is a practical blueprint for keeping data local, training proprietary models, and deploying securely within internal networks.
APEX-S100: Dual-GPU, Short-Depth AI Server for the Edge
Also unveiled: the APEX-S100, a 2U, 420mm-deep short-depth server optimized for edge AI infrastructure.
-
Supports up to two double-width GPUs
-
Modular memory expandability
-
Optimized thermal profile for in-rack deployments with constrained space
This server squarely targets real-time video analytics, AI inference at scale, and ruggedized deployment zones like factories, remote hubs, and autonomous fleets.
Industrial Vision: Modular Camera Systems with GMSL2, USB, and MIPI
Understanding that AI begins with vision, Innodisk is expanding its camera portfolio with industrial modules that support:
-
USB, MIPI, and GMSL2 interfaces
-
Custom drivers and adapter boards for Raspberry Pi, Jetson, and server platforms
-
Real-world demo with Advantech’s TREK-50N and REX-52N systems, showing 16-channel video capture for heavy vehicle surveillance and driver behavior monitoring
The takeaway: Innodisk’s camera architecture is modular, validated, and ready to scale from IoT boards to full AI rigs.
Qualcomm-Powered Smart Parking AI
Partnering with Qualcomm Technologies, Innodisk presented a Smart Parking Recognition and Security System built on the Qualcomm Dragonwing™ IQ9 platform.
-
Leverages an NPU for multi-modal inference (including VLMs and behavior analysis)
-
Performs cloud-independent vehicle recognition (make, model, color) and event tracking
-
Includes camera expansion, local video processing, and security-grade performance
For smart city developers, this is a deployable, edge-native AI system with high-throughput vision processing and zero dependency on the cloud.
Intel Ultra Series 2 Reference Kit
Also highlighted: Innodisk’s compatibility with the latest Intel® Core™ Ultra Series 2 processors.
-
Bundled with MIPI-over-Type C cameras, LPDDR5X DRAM, PCIe Gen5 SSDs, and scalable I/O options
-
A reference kit built for OEMs and integrators seeking next-gen edge AI computing with maximum I/O flexibility
It’s part of a broader effort by Innodisk and Intel to offer turnkey industrial AI packages for surveillance, automation, and data acquisition at the edge.
Memory and Storage for the AI Era
Innodisk continues to innovate in its core competencies—storage and memory—with products now aligned to hyperscale and AI training use cases:
-
DDR5 MRDIMM and LPDDR5X CAMM2 modules for high-density, low-power systems
-
PCIe Gen5 SSDs in U.2, E1.S, E3.L, and E3.S form factors
-
Fully OCP SSD 2.0 compliant, ideal for enterprise AI racks and data centers
These form factors are especially relevant for OEMs standardizing on modular, hot-swappable hardware in edge datacenters, autonomous vehicles, and telco nodes.
Final Thoughts
COMPUTEX 2025 isn’t just about showing speed—it’s about showing solutions. Innodisk brought working demos, validated platforms, and a roadmap that reflects the real needs of integrators building for smart infrastructure, rugged edge environments, and local AI deployment. In a world chasing AI performance, Innodisk quietly delivers AI readiness—across platforms, industries, and form factors.
