The $11 million XPRIZE Wildfire competition closed on September 23 without a winner of either grand prize, as finalists proved they can spot fires from orbit and reach them with drones inside 10 minutes but still can't reliably put them out. The four-year contest split $11M across two tracks: a space-based detection challenge tested in New South Wales in April 2026, and an autonomous response challenge tested across 386 square miles of Nenana, Alaska in June 2026. The $3.5M grand prize from the Minderoo Foundation went unclaimed in the detection track, and the parallel grand prize in the response track went unawarded for the same reason — none of the finalists finished the job.
Anduril Industries took the top autonomous response award of $1.2M, plus an extra $1M from Lockheed Martin for the fastest detection. Its system pairs a Sentry Tower running AI-driven computer vision on thermal imagery with a swarm of surveillance drones, then dispatches a Ghost-X autonomous drone to drop fire-retardant balls on the target. All three finalists in the track detected the high-risk fire in under 10 minutes; two of them launched suppression drones that struck it. None extinguished the blaze.
The detection track drew 45 qualified technical submissions, while the response track pulled in more than 300 teams globally at the start. XPRIZE program director Andrea Santy said early submissions ranged from sound-based suppression to lasers and robotic dogs before judges narrowed to systems that could plausibly hit the 10-minute window. The 15-minute average lag on 911 calls — still the most common way wildfires get reported — is the baseline the competition was trying to beat.
“A 911 call is still the most typical way that we find out about wildfires, and on average, that's going to be at least 15 minutes.”— Andrea Santy, Program director of XPRIZE Wildfire
Key facts
- 01Neither $3.5M grand prize was awarded; no team fully met the 10-minute detect-and-suppress requirement.
- 02Anduril Industries won $1.2M for first place in the autonomous response track, plus $1M from Lockheed Martin for fastest detection.
- 03Dryad Networks took $800K for second, deploying 170 solar ground sensors and a Silvaguard drone in a 386-square-mile Alaska test zone.
- 04SIRIUS Wildfire Alliance won $500K in the space-based track using optical and radar satellite data for wildfire spread modeling.
- 05The competition drew 300+ teams into the response track and 45 qualified submissions for space-based detection.
In the space track, UK-based SIRIUS Wildfire Alliance won first place and $500,000 for a system that combines optical and radar satellite data to identify fires and model their spread. Arizona's Team Snuffed took second and $250,000 for what judges called transparent, auditable, uncertainty-aware detection built for firefighter workflows. Special recognition prizes of $100,000 each went to MyRadar in Florida, DeepFire in Louisiana, and Germany's Mayday.AI, which adapted an algorithm originally built for volcanic ash detection to spot wildfire smoke through cloud cover.
The satellite layer is thin by design. Germany's OroraTech launched a dedicated wildfire detection constellation in March 2025, and Google-backed FireSat put up its first microsatellites in July 2026. Most orbital assets aimed at wildfires were built for weather or general environmental monitoring, so the software task is squeezing wildfire-grade signal out of data that wasn't collected with wildfires in mind. That's where AI carries the load — filtering false positives and turning raw imagery into something a fire crew can act on.
Second place in the autonomous response track, worth $800,000, went to Germany's Dryad Networks. Dryad deployed 170 solar-powered ground sensors that sniff for gases from smoldering fires, then sent an observation drone to geolocate the source before dispatching its Silvaguard Suppression Drone. Third place and $500,000 went to AURA Foresight, a UK-Australia team that co-lead Rob Vernon described as a "Home Depot approach" — off-the-shelf DJI drones coordinated by custom AI software. The team assembled and met in person for the first time in Alaska.
The test itself was harder than a demo. Finalists had to distinguish a high-risk fire from decoys across a rural area the size of the San Francisco Bay Area or San Antonio, Texas, then get suppression payload on target within 10 minutes of ignition. Detection at that speed is now demonstrably possible with the right sensor mix and models. Delivering enough suppressant, fast enough, is the remaining wall.
The competition launched in April 2023 and its relevance has only sharpened since. The Lahaina fire on Maui killed at least 102 people in August 2023; Canada's 2023 season pushed smoke across US cities for weeks. The January 2025 Los Angeles wildfire destroyed nearly 12,000 homes, killed 31 people, and drove an estimated $140 billion in losses — ranking in the top five costliest natural disasters, per Nature Reviews Earth & Environment. South Korea recorded its largest wildfire ever in March 2025, and the 2026 European season forced more than 320,000 people to evacuate in Spain and France while threatening NASA's Deep Space Communications Complex near Madrid.
The unclaimed grand prizes are the honest signal here. Detection has crossed a threshold — orbital constellations, thermal towers, and gas sensors backed by computer vision can find a fire before a human calls it in. Autonomous suppression at the payload weights a drone can carry is still short of extinguishment on anything but the smallest ignitions. Scaling suppressant delivery — larger drones, drone swarms, faster reload cycles, or hybrid deployments with crewed aircraft — is where the next dollar of R&D matters most.
For the AI industry, XPRIZE Wildfire is a useful reminder that the hard part of applied AI is rarely the model. Anduril, Dryad, and SIRIUS built credible detection stacks on existing sensors and satellite data; the bottleneck is physical actuation in the world. Expect defense-adjacent contractors like Anduril and Lockheed to pull further ahead in wildfire response over the next 24 months, because they already own the drone airframes, autonomy stacks, and government procurement channels that municipal fire agencies don't. The detection winners will likely end up as data suppliers into those stacks rather than standalone businesses.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
