Advantech Jetson Thor Series — NVIDIA AI PC
Three Advantech products based on NVIDIA Jetson Thor (T5000 128 GB / T4000 64 GB, up to 2070 FP4 TFLOPS): the MIC-742-AT box system (QSFP28 4 × 25GbE, 8 × GMSL2, 4 × CAN FD), the MIB-742-AT 180 × 180 mm single-board computer (same board, for OEMs) and the AIR-075 industrial box (4 × 10GbE, HDMI + DP, DB9 CAN/DIO/COM). Blackwell GPU, JetPack 7, 19 – 36 VDC.
Manufacturer: Advantech
AI module: Jetson T5000 128 GB (2070 TFLOPS FP4) · T4000 64 GB (1200 TFLOPS)
Network: MIC/MIB: QSFP28 4 × 25GbE + 5GbE · AIR-075: 4 × 10GbE RJ45 (PoE option)
Camera / vehicle: MIC/MIB: 8 × GMSL2 mini FAKRA, 4 × CAN FD · AIR-075: 2 × CAN DB9, 8-bit DIO, 2 × COM
Number of models: 3 (MIC-742-AT, MIB-742-AT, AIR-075)
MIC-742-AT, MIB-742-AT and AIR-075 are Advantech’s AI inference products built on the NVIDIA Jetson Thor platform. With the Blackwell GPU architecture and Arm Neoverse V3AE cores, Jetson Thor delivers 2070 sparse FP4 TFLOPS and 128 GB LPDDR5X on the T5000 module and 1200 TFLOPS and 64 GB on the T4000; Advantech positions the platform for large language and vision-language models, generative AI, multi-camera perception and physical AI (robotics, autonomous machines). MIC-742-AT is the MIC-AI series box system with QSFP28 25GbE and GMSL2 cameras; MIB-742-AT is the enclosure-less OEM version of the same board; AIR-075 is the AIR line’s industrial box computer with 10GbE RJ45 and DB9 interfaces and Edge AI SDK support. FST presents the three as the Thor grouping in the NVIDIA AI PC category; each model page has the full datasheet-based specification table, part numbers and accessories.
Common platform: Jetson Thor T5000 / T4000
On all three products the module is selected by part number: T5000 (14-core Neoverse V3AE, 2560 Blackwell CUDA + 96 fifth-gen Tensor Cores, 128 GB LPDDR5X at 273 GB/s) or T4000 (12-core, 1536 CUDA + 64 Tensor Cores, 64 GB). The operating system is NVIDIA JetPack 7 (MIC/MIB 7.0, AIR-075 Ubuntu 24.04 + JetPack 7.1); M.2 2230 E key Wi-Fi and M.2 3052 B key 5G slots, M.2 NVMe storage, TPM and the 19 – 36 VDC power input are common. Advantech’s MIC-741-AT (SFP28 + 2 × 10GbE, 2 × isolated CAN FD), MIC-743-AT (desktop box) and MIB-741-AT (T4000) are preliminary; no datasheet is published for AIR-073 (T3000 / T2000, 865 TFLOPS).
Which model for which job
MIC-742-AT (217 × 76.4 × 184.5 mm, 3 kg): for high-bandwidth perception and robotics — four (three on T4000) 25GbE from the QSFP28 cage, 5GbE RJ45, four CAN FD, I2C, four USB 3.2 Gen 2, two SATA and optional 8-ch GMSL2 mini FAKRA camera input; wall or DIN-rail mount, −10 to 60 °C. MIB-742-AT (180 × 180 mm, 1 kg): the SBC form of the same board with module — for machine and vehicle builders designing their own enclosure; the 130 W cooler and an external adapter for the GMSL board are optional. AIR-075 (235.8 × 174 × 59 mm, 2.89 kg): industrial box-PC layout — four (two on T4000) 10GbE RJ45 (PoE option), HDMI 2.1 + DP 1.4a, CAN / DIO / two COM on DB9, two USB-C + two USB-A, two M.2 NVMe; for factory and field installations with Edge AI SDK / Inference Kit / SUSI.
Cameras, network and wireless
On MIC/MIB, cameras connect over GMSL2 to the MIC-FG-8G2C1 frame grabber (mini FAKRA, 8 channels); on AIR-075, GigE Vision cameras connect to the 10GbE ports (PoE with the optional MIOe PSE module) and USB 3.2 Gen 2 ports serve USB3 cameras. For the network backbone, QSFP28 25GbE is used on MIC/MIB and 10GbE on AIR-075. Wireless: M.2 E key Wi-Fi 6 / 6E and M.2 B key 5G + GNSS modules (nano SIM); antenna and cable part numbers are on the model pages. Storage: 1 TB NVMe is installed on the MIC/MIB A1 SKUs and 512 GB on the AIR-075 SSD SKUs; SQFlash and SSSTC options go up to 4 TB.
Where it is used
Multi-camera visual inspection and generative AI assistants on the factory floor, perception/planning computers for humanoid and mobile robots, autonomous machinery and vehicles (CAN FD, GMSL2), smart-city and traffic analytics nodes, on-premises large language and vision-language model inference, medical and laboratory imaging workstations. FST delivers the Thor system together with the camera / sensor layer, 25G / 10G network design, model deployment (JetPack 7, Edge AI SDK) and commissioning services as one solution.
Models in the series
The three models of the Thor grouping are listed below; click a model name in the cards or the table for specifications, ordering and accessory information. Contact us for MIC-741-AT, MIC-743-AT, MIB-741-AT (preliminary), AIR-073 (no datasheet) and the IGX based MIC-735 / MIB-735 in the same Advantech line.
MIC-742-AT
T5000 / T4000 · 2070 TFLOPS · QSFP28 25GbE · 8 × GMSL2 · 4 × CAN FD · In production
MIB-742-AT
T5000 / T4000 · 2070 TFLOPS · 180 × 180 mm SBC · QSFP28 · 8 × GMSL2 · New
AIR-075
T5000 / T4000 · 2070 / 1200 TFLOPS · 4 × 10GbE · HDMI + DP · 2 × CAN · New
| Model | Jetson module | Interfaces | Dimensions | Mounting / power |
|---|---|---|---|---|
| MIC-742-AT | Jetson T5000 128 GB / T4000 64 GB | QSFP28 4 × 25GbE · 5GbE · 4 × USB 3.2 · HDMI · 4 × CAN FD · 8 × GMSL2 | 217 × 76.4 × 184.5 mm | Wall / DIN rail · 19 – 36 VDC |
| MIB-742-AT | Jetson T5000 128 GB / T4000 64 GB | QSFP28 4 × 25GbE · 5GbE · 4 × USB 3.2 · HDMI · 4 × CAN FD · 8 × GMSL2 | 180 × 180 mm | SBC, wall · 19 – 36 VDC |
| AIR-075 | Jetson T5000 128 GB / T4000 64 GB | 4 × / 2 × 10GbE · HDMI + DP · 4 × USB · 2 × CAN · 8 bit DIO · 2 × COM | 235.8 × 174 × 59 mm | Wall · 19 – 36 VDC |
Which Jetson Thor system is right for your project?
Share your application (robotics / visual inspection / generative AI), the number and type of cameras (GMSL2 / GigE / USB), the network needs (25GbE / 10GbE), the CAN and I/O requirements, your enclosure preference (box / SBC) and the ambient temperature; let’s decide on the module (T5000 / T4000), the model and the accessory set together.
All product specifications are subject to change without notice; data compiled from Advantech’s MIC-742-AT (14 August 2026), MIB-742-AT (26 December 2025) and AIR-075 (25 August 2026) datasheets and the advantech.com product pages.
Source: Advantech product data