RICHARDSON, TX — Amazon’s ambitious quest to dominate the low-altitude logistics market has encountered another significant setback after one of its next-generation MK30 drones crashed into an apartment complex in North Texas. On February 6th, 2026, federal investigators and company engineers are scrutinising the wreckage of the incident, which marks the second high-profile failure for the Prime Air program in the state within a three-month window.
Residents in the Dallas-Fort Worth suburb of Richardson were startled earlier this week when the hexacopter, designed for high-precision autonomous delivery, veered off course. Resident Cessy Johnson captured the descent on video, filming as the aircraft moved erratically toward the side of a residential building.
“The propellers on the thing were still moving, and you could smell it was starting to burn,” Johnson told local news outlets, noting that she had begun recording simply because she had never seen a delivery drone in action before. “And you see a few sparks in one of my videos.”
While the Richardson Fire Department arrived on the scene shortly after the impact, no blaze was reported, and the building sustained only minor structural damage. However, for an industry striving for public trust and regulatory approval for beyond visual line of sight (BVLOS) operations, the optics of a 50-pound drone striking a multi-family dwelling are damaging.
MK30 Drones
Photo: Amazon News
Technological Failures and the Sense-and-Avoid Gap
The MK30 was touted by Amazon as a revolutionary leap over its predecessor, the MK27-2. It features a quieter acoustic profile and a more sophisticated “sense-and-avoid” suite designed to detect and navigate around wires, chimneys, and even household pets.
Despite these upgrades, the Richardson crash follows a pattern of stationary object collisions. In November 2025, an MK30 severed an internet cable in Waco, Texas, during its ascent. Just weeks prior, two drones were destroyed in Arizona after colliding with the boom of a construction crane. These recurring incidents suggest a potential systemic vulnerability in the drone’s perception stack, the fusion of software and hardware that allows the aircraft to “see” its environment.
Official Response and Accountability
Amazon has moved quickly to contain the fallout, offering to repair the building and issuing a formal apology. In a statement provided to FLYING Magazine and other aerospace journals, an Amazon spokesperson addressed the Richardson event:
“We apologise for any inconvenience and are actively investigating the cause of this incident.”
The Federal Aviation Administration (FAA) has been notified and is expected to integrate this latest crash into its ongoing probe of Prime Air’s safety protocols. While the National Transportation Safety Board (NTSB) has not yet opened a formal investigation into the Richardson strike, it continues to monitor the “non-action” events that are beginning to cluster around the MK30’s deployment.
Photo: Amazon News
Amazon Prime Air Incident Log 2025-2026
The following table tracks the most recent significant incidents involving the Amazon MK30 fleet as the company pushes toward its goal of 500 million annual deliveries by 2030:
DateLocationIncident TypeImpact / ResultStatusOct 2025Tolleson, AZCrane Collision2 Drones destroyed; battery fireFAA/NTSB InvestigationNov 18, 2025Waco, TXCable StrikeInternet line severed; “safe” descentFAA ProbeDec 2025Pendleton, ORSoftware Failure2 Drones crashed during testingInternal ReviewFeb 4, 2026Richardson, TXBuilding StrikeProperty damage; sparking wreckageActive Investigation
The Path Forward for Prime Air
Despite the turbulence, Amazon is not grounding the fleet. The company recently received Civil Aviation Authority (CAA) approval to begin trial flights in Darlington, UK, and is continuing to add delivery sites in Michigan and Missouri. Analysts suggest that Amazon is willing to absorb the cost of these “learning moments” in a closed-loop testing environment, but as these incidents move from private ranges to residential backyards, the tolerance for error from both regulators and the public is rapidly shrinking.
The core challenge remains the “thin-wire” and “complex-background” problem. While the MK30 can successfully avoid a parked van or a large tree, detecting a non-reflective utility cable or the exact corner of a brick building in varying light conditions remains a formidable hurdle for current machine learning models.