NVIDIA Jetson T4000 Edge AI embedded system offers 5GbE networking, four PoE camera ports, DIO, CAN Bus, and more

AAEON BOXER-8742AI is a fanless Edge AI embedded system powered by an NVIDIA Jetson T4000 that was unveiled along with the Jetson Thor T5000 module last August, but officially launched at CES 2026.

The Jetson T4000 is a weaker variant, but it’s still extremely powerful with up to 1200 TFLOPS of AI computing power against 2070 TFLOPS for the T5000. The BOXER-8742AI features a 5GbE port, four Gigabit Ethernet RJ45 PoE ports suitable for IP cameras, DIO, RS232, RS485, and CAN Bus interfaces, four USB 3.2 Gen 2 ports, HDMI video output, and more. It’s suitable for high-performance edge AI applications such as smart factory automation, AMR, and AI-assisted roadside unit solutions.

AAEON BOXER-8742AI specifications:

• System-on-Module – NVIDIA Jetson T4000
  • CPU – 12-core 64-bit Arm Neoverse-V3AE
  • GPU – 1536-core NVIDIA Blackwell GPU with 64 Tensor Cores, 6-TPC MIG
  • VPU
    • Video Decode – 4x 8Kp30 (H.265), 10x 4Kp60 (H.265), 22x 4Kp30 (H.265), 46x 1080p60 (H.265), 92x 1080p30 (H.265), 82x 1080p30 (H.264), 4x 4Kp60 (H.264)
    • Video Encode – 6x 4Kp60 (H.265), 12x 4Kp30 (H.265), 24x 1080p60 (H.265), 50x 1080p30 (H.265), 48x 1080p30 (H.264), 6x 4Kp60 (H.264)
  • AI performance – 1200 TFLOPS FP4 (Sparse)
  • Memory – 64 GB 256-bit LPDDR5X up to 273 GB/s
• Storage – M.2 2280 Key-M slot for NVMe SSD
• Video Output – HDMI 2.0 x 1
• Networking
  • 5GbE RJ45 port
  • 4x Gigabit Ethernet RJ45 with PoE 802.3af/at (up to 60 Watts)
  • Optional WiFi and Bluetooth via M.2 E-Key module
  • Optional 4GT LTE/5G cellular via M.2 B-Key module + 2 SIM card slots
  • 6x external SMA antenna openings
• USB
  • 4x USB 3.2 Gen 2 (Type-A) port
  • Mini USB port for OS flash
• I/Os
  • RS-232/422/485 and 8x  DIO via DB-15 connector
  • RS-232 and 2x CAN FD via DB-9 connector
• Expansion
  • M.2 2280 M-Key socket for NVMe  SSD
  • M.2 3042/3052 B-Key socket (USB 3.0 + PCIe) + 2x NanoSIM slots for 4G LTE/5G cellular connectivity
  • M.2 2230 E-Key socket for WiFi/Bluetooth/NVMe
• Misc
  • Power, Reset, and Recovery buttons
  • Power LED
• Power Supply – 12V-24V DC input via 2-pin Phoenix connector
• Dimensions – 210 x 164.2 x 74mm
• Temperature  Range
  • Operating  – -25°C to +60°C, according to IEC60068-2 with 0.7 m/s airflow
  • Storage – -40°C to +85°C), non operation
• Humidity – 5%~ 95% @ 40°C, non-condensing (storage)
• Anti-Vibration – 3.5G / 5~500Hz / operating
• Anti-Shock – 50G peak acceleration ( 11 msec.)
• Certifications – CE/FCC class A

The fanless Edge AI embedded computer is supported by the NVIDIA Jetpack 7.0 SDK (or greater), like most other Jetson modules.

As with almost all products announced at CES 2026, the Jetson Thor T4000 is not available just yet, and nor is the BOXER-8742AI, which is expected to be available in Q2 2026. What we know do know is that the NVIDIA Jetson Thor T4000 system-on-module itself will sell for $1,999 at 1,000-unit volume, so the AAEON unit will likely be around $3,000. A few more details may be found on the product page, including a preliminary datasheet, but not much else.

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress. We also use affiliate links in articles to earn commissions if you make a purchase after clicking on those links.