There are more companies selling robot intelligence right now than companies selling robots.
That showed up this year as a pattern I kept noticing.
In this case all these research groups with published on the same robot from
@DexmateAI.
DeepMind’s Gemini Robotics 2 launch video. Amazon FAR’s real-world RL on a dexterous humanoid. DUET, the dual-robot manipulation framework from USC, Stanford and TRI.
A 100-hour tactile dataset from Berkeley, NVIDIA and Stanford. World Labs’ Atlas launch. And a Mayo Clinic residency track, which is not a robotics lab at all.
The platform in all of them is Vega, from Dexmate.
Models, VLAs, data engines, simulators, teleop stacks: that side of the market is crowded.
The side that will sell a lab a capable bimanual mobile robot, this quarter, at a published price, is thin.
So a team with a policy to validate either waits on a partnership or spends a year building a platform that was never the point of the paper.
Dexmate sells the platform. Look at what the spec sheet commits to, and what it refuses:
→ Vega 1 Pro, $75,000 at this year’s pricing, roughly four weeks from order
→ 24 DoF, torso from floor to 2.2 m, 5 kg per arm, omni base, ~20 h per charge
→ Jetson AGX Thor, stereo head, two LiDAR, four base cameras
→ Python, C++ and Rust APIs in the box
→ SimReady URDF and USD models matching the shipped robot
Not the model company, not the data company, not the simulation company. The thing they run on.
Congrats
@taochenshh,
@qinyuzhe and the team! :)
~~
♻️ Join the weekly robotics newsletter, and never miss any news →
ziegler.substack.com