- Nvidia’s new Jetson AGX Thor Developer Kit delivers up to 2070 TFLOPS of performance in a compact form.
- Powered by the Jetson T5000 module with a Blackwell GPU, 14-core CPU, and 128GB of LPDDR5X RAM.
- Designed for robotics, AI vision, and sensor-heavy workloads, with support for multiple 8K video streams.
- Available now for pre-order at $3,499, with shipments starting November 20, 2025.
Nvidia has officially unveiled what may be its most ambitious small-form-factor computing platform to date. The new Jetson AGX Thor Developer Kit is a workstation-class mini PC that combines immense GPU power with a compact design aimed squarely at researchers, robotics engineers, and AI developers.
With a staggering 2070 TFLOPS of AI performance, the Jetson AGX Thor pushes the Jetson family into a completely new performance category. For context, that level of throughput was unthinkable in a system of this size just a few years ago.
Nvidia’s move signals how rapidly compact AI systems are evolving and how much emphasis the company is placing on edge AI and robotics computing.
What powers the Jetson AGX Thor
At the heart of this tiny powerhouse is the Jetson T5000 system-on-module, built using Nvidia’s latest Blackwell GPU architecture. This GPU packs 2,560 CUDA cores, 96 fifth-generation Tensor Cores, and leverages Multi-Instance GPU (MIG) capabilities to let developers partition workloads across applications with maximum efficiency.
On the CPU side, Nvidia has paired the GPU with a 14-core Arm Neoverse-V3AE processor, providing the versatility to handle both traditional workloads and AI-heavy applications. This combination ensures that the Jetson AGX Thor isn’t just about raw GPU horsepower but is equally adept at juggling diverse, real-world computational tasks.
Backing this up is a generous 128GB of LPDDR5X memory. For a system of this size, that is enormous. Memory has often been the bottleneck in smaller developer systems, and Nvidia’s choice here highlights that Thor is designed to push the boundaries of what edge systems can handle.
Designed for vision, robotics, and sensor-heavy workloads
The Jetson AGX Thor is not your typical mini PC for casual computing. It is designed for advanced vision-based applications. Nvidia says the system can simultaneously encode and decode multiple streams of 4K and even 8K video, a critical capability for robotics, AI vision systems, and advanced industrial workflows.
This is also why connectivity has been given serious attention. The system includes four 25GbE channels for high-speed networking, along with PCIe support for NVMe storage. Out of the box, the developer kit ships with a 1TB NVMe SSD, a WiFi 6E module, HDMI 2.0b and DisplayPort 1.4a outputs, and an assortment of USB ports. For expansion, developers get M.2 slots, making it easier to add sensors, storage, or custom modules.
Its compact footprint of 243 x 112 x 56 mm makes it significantly smaller than most workstations. While it’s bigger than an ultra-compact office PC, it is tiny compared with the powerful computing hardware it rivals.
This balance between size and performance makes it attractive for engineers working on humanoid robots, autonomous machines, and AI systems that need processing power close to the action.
A more affordable sibling: Jetson T4000
Alongside the flagship T5000, Nvidia also introduced the Jetson T4000, a lighter and more affordable option that is still in development. While not as powerful as its big brother, the T4000 is no slouch. Early specifications reveal 1,200 TFLOPS of FP4 performance, a 1,536-core GPU, and 64GB of LPDDR5X memory.
The T4000 has a power range of 40 to 75 watts, compared with the T5000’s 40 to 130 watts, making it a more energy-efficient choice for developers who don’t need maximum performance. This could appeal to robotics startups, researchers, or companies testing AI models where efficiency and cost matter more than raw throughput.
Why Nvidia is pushing compact AI systems
Nvidia is positioning the Jetson AGX Thor Developer Kit as part of a broader strategy to support AI innovation beyond the data center. While its DGX systems remain the heavy hitters in enterprise and research labs, the Jetson family is designed for edge AI applications where computing needs to happen closer to the physical environment.
Thor in particular is being pitched for humanoid robotics, sensor-driven applications, and complex vision AI systems. Nvidia’s existing software ecosystems — including Isaac for robotics, Metropolis for smart cities, and Holoscan for healthcare AI — will tie directly into the new platform, giving developers a full-stack solution from hardware to software.
By introducing such a powerful edge computing system, Nvidia is also giving smaller labs, startups, and independent researchers access to performance levels that were previously available only in bulky, expensive systems. With this move, the company is essentially shrinking a supercomputer into a compact box.
Price, availability, and what comes next
The Jetson AGX Thor Developer Kit is priced at $3,499, positioning it as a premium system for professional developers and organizations. Pre-orders are already live through select distributors, with shipments scheduled to begin on November 20, 2025.
While $3,499 is not cheap, the performance-to-price ratio is compelling for researchers who might otherwise need a full workstation or cloud compute resources to achieve similar results. For organizations deploying AI at the edge, the Thor could represent a cost-effective, high-performance alternative to traditional setups.
The Jetson T4000, meanwhile, has not yet received final pricing or a firm release date, but it is expected to arrive sometime after the T5000 launches.
Final thoughts
With the launch of the Jetson AGX Thor, Nvidia is once again reshaping what’s possible in the world of compact computing. Delivering 2070 TFLOPS in a mini PC-sized package is a remarkable engineering achievement, and it underscores how critical edge AI is becoming across industries.
From humanoid robots to AI-powered video analysis, Nvidia is making it clear that high-performance AI doesn’t need to live only in the cloud or data center. It can now sit on a developer’s desk, or even inside a robot.
For engineers, researchers, and companies pushing the limits of AI innovation, Thor is not just another addition to the Jetson family. It’s a glimpse into the future of small but incredibly powerful machines.
Follow TechBSB For More Updates
