Joby Aviation’s AI-Optimized Rotors: A Masterclass in Aerodynamic Supercomputing

A six-rotor aircraft lifted itself off the ground, tilted forward, flew like a plane, and then settled back down like a helicopter last week at Marina Airfield, where there is a specific type of California morning when the fog lingers long enough to soften everything. In the cockpit was James “Buddy” Denham. Engineers have been waiting years for this kind of moment, which doesn’t really show up until much later.

Since 2009, Joby Aviation has been developing this aircraft. By Silicon Valley standards, that’s a long runway, and the patience has begun to seem almost archaic. Joby consistently shows up with something fresh to show, while other eVTOL contenders have been quietly folding, running out of money, or losing the trust of their supporters. The contrast is difficult to ignore.

The S4’s rotors have a more fascinating technical history than the press releases suggest. Each of the six rotors, which alternate between vertical and horizontal flight, is controlled by software that must think more quickly than a human pilot could. This is the point at which the NVIDIA partnership becomes significant. The new NVIDIA IGX Thor platform, which is based on the Blackwell architecture and is essentially industrial-grade compute intended for physical AI applications, has chosen Joby as its sole aviation launch partner. The aircraft now has a brain that can process enormous amounts of sensor data in real time, to use marketing jargon.

Joby’s Flight Research Lead, Gregor Veble Mikić, made a comparison that sticks. He noted that autonomous vehicles have already demonstrated that computers are capable of interpreting massive amounts of data to make snap decisions. However, the bar is higher for an aircraft. Every computation needs to be flawless. Every choice is perfect. At 5,000 feet, there is no shoulder to stop on.

Radar and perception processing, sensor fusion, autonomous mission management, and predictive system health monitoring are all planned features of the SuperpilotTM system. Additionally, a “digital twin”—a virtual representation of the aircraft that learns from each flight and improves over time—is being discussed. It sounds futuristic until you consider that offshore wind farms and Formula 1 garages are already experiencing this kind of thing. The aviation industry is merely catching up, and doing so with caution.

Compared to its rivals, investors appear to think Joby is closer to commercial reality. The FAA certification process is progressing, despite the ambitious and perhaps optimistic plan to begin service in Dubai later this year. People tend to forget how much the early Model S was ridiculed before it wasn’t, and Tesla faced similar doubts years ago. Joby seems to be in a similar stage right now, having moved past the stage where critics refer to it as vaporware but not quite to the point where it is well-known.

Whether the AI-optimized rotor system will withstand the chaotic, unpredictable conditions of real urban airspace is still up in the air. Marina, California is a controlled setting. The skies in Dubai, New York, and Los Angeles will test different things. As this develops, it’s easy to imagine Joby as the business that eventually makes flying taxis a reality. The most difficult part may still lie ahead.

The rotors are spinning in any case. Additionally, the eVTOL discussion feels, for the first time in a long time, less like a pitch deck and more like the slow, deliberate penmanship of aviation history.