What technologies helped NVIDIA win four COMPUTEX 2026 Best Choice Awards, and how do the Jetson Thor edge AI module, the Vera Rubin NVL72 rack scale AI supercomputer, and the Alpamayo autonomous driving platform improve AI computing, robotics, and self‑driving systems in terms o

How Jetson Thor improves robotics:
Its value is not just raw speed; it is the combination of higher compute, tighter power/thermal operation, and local real-time reasoning for industrial systems, humanoids, and other robots that need fast decisions from camera and sensor streams.

Vera Rubin NVL72 rack-scale AI supercomputer:
Vera Rubin NVL72 is a rack-scale system that combines 36 Vera CPUs and 72 Rubin GPUs connected with sixth-generation NVLink so the GPUs can operate as one large AI system for trillion-parameter training, post-training, inference, and long-context workloads.
NVIDIA says it can train AI with one-fourth the GPUs of Blackwell-class systems and deliver AI inference at one-tenth the cost per million tokens, with up to 10x higher inference throughput and up to 10x better inference performance per watt in cited platform material.

How Vera Rubin improves AI computing and future infrastructure:
Its main infrastructure advance is that AI performance is moving from a chip race to a rack- and factory-scale systems race, with compute, interconnect, cooling, and networking engineered together.
NVIDIA positions Vera Rubin as infrastructure for agentic AI and reasoning workloads, especially trillion-parameter models and million-token context windows, which are important for future AI factories and large-scale inference services.

Alpamayo autonomous driving platform:
Alpamayo is an open platform for autonomous-vehicle development and was recognized in the smart vehicle cockpit category at COMPUTEX.
Available reporting says Alpamayo is designed as a visual-language-action driving model with reasoning capabilities, generating driving trajectories along with reasoning traces to make decisions more explainable in difficult road scenarios.

How Alpamayo improves self-driving systems:
The reported benefit is a shift from perception-only systems toward models that can reason about what they see and why they choose a maneuver, which can help with edge cases and transparency.
Reporting also says Alpamayo training combines real-world driving data with simulation/synthetic data, which matters because self-driving systems need broader coverage than road testing alone can provide.

Best overall interpretation:
Jetson Thor improves edge AI and robotics with much higher local compute and better efficiency.
Vera Rubin NVL72 improves centralized AI infrastructure with rack-scale performance, lower token cost, and better performance per watt for large reasoning models.
Alpamayo improves autonomous driving software by adding reasoning-oriented models trained on broader driving data, including real-world inputs.

Evidence note:
The awards and the Jetson/Vera Rubin technical claims are well supported by NVIDIA and event coverage.
The Alpamayo details are less directly documented in the available official snippets I found, so its reasoning and mixed-data training description relies more on secondary reporting than on a clearly surfaced primary technical document.