YUAN begins Jetson T3000 and T2000 platform development.

YUAN High-Tech has begun developing edge AI platforms based on the NVIDIA Jetson T3000 and T2000 modules. Both use the NVIDIA Blackwell architecture and the same software family as the Jetson AGX Thor T5000, giving developers a common starting point across the three modules.

YUAN already offers T5000 platforms for factory inspection, autonomous robotics, and on-site video analysis. The T3000 and T2000 development work adds options for projects with different performance, power, size, and unit-cost constraints.

T3000: lower memory and power than T5000

NVIDIA's announced specifications position the Jetson T3000 at roughly half the T5000's memory capacity, power consumption, and module size, with around 90 percent of its AI inference performance. It uses the same NVIDIA software family, which may help teams target a smaller system while retaining a related development environment. The IGX T3000 also retains the Functional Safety Island for projects with functional-safety requirements.

T2000: Blackwell compute at 40 W

NVIDIA lists the Jetson T2000 with 400 FP4 TFLOPS of AI compute, a Blackwell GPU, 16 GB of LPDDR5X memory, and a 40 W power envelope. Those specifications make it a candidate for visual AI, mobile robots, robotic arms, and other edge systems with tighter power or cost limits.

Emulation schedule

The modules are expected in Q1 2027. NVIDIA plans T3000 emulation with JetPack 7.2.1, allowing customers to test applications on an existing Jetson AGX Thor development platform. T2000 emulation support is planned for a later release.

The planned T3000 and T2000 systems will join YUAN's existing T5000 platforms. Final module availability, production specifications, and migration requirements should be confirmed before project planning.