NVIDIA Jetson Thor targets advanced humanoid apps, driving humanoid robot chip market to exceed $48 million by 2028

NVIDIA’s newly launched Jetson Thor acts as the “physical intelligence core” of humanoid robots, integrating the Blackwell GPU and 128 GB of memory to deliver 2,070 FP4 TFLOPS of AI performance—7.5 times that of its predecessor, Jetson Orin. 

TrendForce reports that this leap is not just a numerical upgrade; it enables robots to process massive sensor data and LLMs in real time, allowing advanced humanoids to truly see, reason, and act. With companies such as Agility Robotics, Boston Dynamics, and Amazon adopting the platform and building ecosystems, the humanoid robot chip market is projected to surpass US$48 million by 2028.

While hardware and software innovations continue, the key driver of volume adoption remains the expansion of real-world application scenarios. In its late-2025 report, the International Federation of Robotics (IFR) observed that while countries are taking different development paths, humanoid robot adoption will generally follow a staged progression.

Near-time deployments will remain limited to pilot projects, mid-term adoption will scale into manufacturing and service sectors, and long-term expansion will extend into everyday household use, where high-performance SoCs will be essential. Similarly, TrendForce’s “1H25 In-Depth Robotics Report” projected that global humanoid shipments will not surpass the 100,000-unit milestone until around 2032, which is consistent with the IFR’s phased outlook.

Vendors are stepping up investments as near- and mid-term use cases grow clearer. At the recent Taiwan Robotics and Automation Show, Advantech, NEXCOM, and Solomon showcased products and services built on Jetson Thor.

Taiwanese suppliers are also moving beyond single components such as servo motors, precision reducers, and advanced sensors to provide integrated solutions. Some are even introducing modularized offerings, likely easily assembled vision and sensing modules, which reduce entry barriers and enhance competitiveness in global supply chains.

TrendForce points out that despite its performance advantages, the Jetson Thor development kit is priced at $3,499—more than double the $1,499 price tag of the Jetson Orin. Given industry efforts to lower humanoid robot costs for broader adoption, vendors focused on simpler tasks may still rely on lower-priced chips.

NVIDIA is expected to leverage its hardware-software integration strategy, releasing dedicated software platforms for Jetson Thor to accelerate development cycles, streamline task execution, and maximize the value of its higher AI compute costs.

Source: TrendForce, Taiwan.