BitRobot Foundation Launches $5M Grand Challenges Fund

For decades, progress in robotics has been locked inside closed labs. Carefully staged demo videos and corporate R&D silos have hidden the real pace of advancement.

We believe the future of embodied intelligence can’t be built that way. Progress has to be open, measurable, and collective.

That’s why the BitRobot Foundation is launching the $5M Grand Challenges Fund—a global prize competition designed to accelerate breakthroughs in robotics and embodied AI.

Inspired by DARPA, Built for Robotics

The program takes inspiration from DARPA’s legendary Grand Challenge, which transformed autonomous vehicles from an academic curiosity into a real industry. DARPA’s simple formula worked: set clear benchmarks, pit human experts against machine intelligence, and make progress measurable by all.

The BitRobot Grand Challenges extend that model to embodied AI. Each category carries up to a $1M prize for the first robotics system to decisively outperform human experts in its domain, proof that autonomy has crossed into the superhuman.

The Inaugural Challenges

Three “moonshot” tasks mark the beginning of this global program. Each is simple to describe, but difficult to master:

Urban Navigation: 

AI models will face off against human pilots in long-range city navigation missions, beginning in the Bay Area before expanding to cities worldwide. Success here could pave the way for safer, more efficient autonomous transportation in dense urban environments.

IKEA Furniture Assembly: 

Robots will attempt increasingly complex multi-part furniture builds, testing manipulation and planning at human levels. Advancements in this challenge translate to real-world applications in logistics, home assistance, and industrial automation.

Origami Dexterity: 

Co-organized with the Nippon Origami Association, this challenge will push fine-motor control, with robots attempting to fold designs that demand the precision and creativity of skilled origami masters. Breakthroughs in fine-motor control here have implications for delicate tasks in fields like surgery, electronics, and caregiving.

Each competition comes with transparent rules, evaluation criteria, and open submission guidelines.

Humans as the Benchmark

Success will be unmistakable. In navigation, embodied AI will be tested against gamers with hundreds of hours of piloting experience. In origami, robots must match the artistry of human masters. A decisive win over these human baselines means the field has reached a new threshold of physical intelligence.

Regular Competition, Open Results

Challenges will run quarterly, often aligned with major academic conferences like ICRA and IROS. Results, datasets, and methods will remain open, giving the global research community shared milestones and reproducible progress.

Driving AI and Hardware Together

Though the program benchmarks AI, it is expected to accelerate advances in hardware as well. Better hands, sensors, and locomotion will enable breakthroughs in manipulation and navigation—capabilities that ripple across the entire field.

An Open Call to Innovators

The Grand Challenges are open to universities, startups, independent researchers, and hobbyists. Anyone can propose new challenges. Sponsors are also invited to expand the prize pool, ensuring that more moonshot tasks can be tackled.

“DARPA’s Grand Challenges reshaped the future of self-driving cars,” said Jonathan Victor, President of the BitRobot Foundation. “With BitRobot, we’re taking that model further. By testing AI against human experts, holding competitions regularly, and sharing results openly, we can chart the real progress of embodied AI and bring robotics closer to everyday impact.”

Why It Matters

The breakthroughs that define the future won’t happen behind closed doors. They will happen in public, where anyone can measure, replicate, and build upon them.

That’s what the BitRobot Grand Challenges are about—breaking robotics open, and proving together when machines truly match or surpass human ability.

Will your machine be the one that does it?