Amazon’s futuristic delivery vision has produced a spectacularly awkward moment after a Prime Air drone dropped a customer’s package directly into a backyard pool in Texas. Video of the incident has rapidly spread online, showing the autonomous delivery aircraft hovering over a residential property before the package falls into the water instead of reaching its intended drop zone. The mishap has turned what was supposed to be a showcase for faster, automated delivery into a viral reminder that even sophisticated delivery systems can make surprisingly basic mistakes.

The incident comes as Amazon is aggressively expanding Prime Air, with the company announcing plans to bring drone delivery to nearly 500 U.S. cities and towns by the end of 2026. Amazon says its MK30 drones can carry packages weighing up to five pounds and use onboard sensing and avoidance technology to navigate their surroundings before releasing a package at a designated location. The company has promoted the system as a faster way to deliver eligible products, with some deliveries designed to arrive within about 30 minutes.

But the pool mishap highlights a problem that is difficult to solve with software alone: the final few feet of a delivery can be just as important as the flight itself. Reports suggest that propeller wash and the challenges of accurately positioning a package can contribute to water-related delivery failures, while reflective surfaces such as pools may create additional complications for sensors. Amazon has also faced other drone incidents, including a February crash involving a Prime Air aircraft that struck a North Texas apartment building, although no injuries were reported in that case.

The viral failure raises an uncomfortable question for Amazon as it prepares to put far more delivery drones into American neighborhoods: how reliable can autonomous delivery become when the final drop is surrounded by pools, ponds, trees, power lines and other obstacles? Amazon argues that safety remains a priority and has invested heavily in autonomous navigation and regulatory testing, but every highly visible mistake gives critics another reason to question whether the technology is ready for mass deployment. As the company moves toward hundreds of service locations, the difference between a futuristic convenience and an expensive package rescue may ultimately come down to whether the drone can identify exactly where it should—and should not—drop the box.