🤖 Innodata Opens Its Own Physical AI/Robotics Lab
On September 30, 2026 , Innodata opened a new motion-capture AI lab in New Jersey . The lab was established in collaboration with Vicon and is capable of tracking human and robotic movements with submillimeter accuracy .
The key point:
- Training data for humanoid robots
- Data for industrial robotics
- Capturing real 3D movements instead of just 2D video analysis
- Customers can purchase ready-made datasets or have custom datasets created
- Robots can be brought into the lab and independently tested for their actual performance be
- Innodata thus offers not only data production, but also evaluation and safety/performance validation .
Why I find this particularly interesting for the 2036 portfolio
This changes Innodata’s story somewhat.
Until now, the investment story has primarily been:
Generative AI → Training data → Data preparation → AI models
Now another layer is being added:
Physical AI → Robots → Real-world motion data → Training → Evaluation → Safe autonomous systems
This fits directly with the Depot 2036 theme “Humanoid / Autonomous Systems”.
And even more interesting: Innodata appears to be attempting to cover the entire data cycle:
Real-World Data → Training → Model Evaluation → Performance Validation
Strategically, this sets it apart from being merely a traditional data labeling provider.
📈 And its operational foundation is also very strong at the moment
The Q2 figures were already exceptional:
Q2 2026 Innodata
Revenue
$92.1 million
Growth
+58% YoY
Adjusted EBITDA
$25.4 million
EBITDA growth
+92%
Adjusted Gross Margin
49%
The company also had, as of the end of June, $250.4 million in cash, including short-term investments, which included significant customer prepayments.
Innodata – Innodata meldet Rekordergebnisse für das zweite Quartal 2026
My Depot 2036 Conclusion:
The new robotics motion capture initiative is not a side issue. It is a concrete step from “AI data” toward “physical AI data infrastructure” —and that is precisely why I will be following this development very closely during Innodata’s upcoming earnings call.
The connection between Innodata and Meta Muse is currently very interesting, but has not yet been officially confirmed.
- According to Hunterbrook, Meta is said to be Innodata’s largest customer. For 2025, revenue is projected to be around $146 million in revenue from Meta.
- In August, Innodata announced a new, scalable program with its largest customer: “personalization of long-horizon agents” —exactly the kind of work that suits an agent like Muse.
- Hunterbrook therefore sees a possible connection to Muse. Neither Meta nor Innodata has confirmed this connection.
What’s particularly exciting: Muse isn’t just any ordinary chatbot, but an autonomous AI agentthat can perform tasks across various apps. Training, evaluation, data preparation, and reinforcement learning environments are particularly important for this—and these are precisely the areas where Innodata specializes.
Ist Muse ein Gewinn für Innodata?
This is not yet a confirmed contract for Muse, but it’s a remarkable chain of circumstantial evidence. If it’s later officially confirmed that Innodata is indeed involved in Muse or its training/further development, that would be a major development for Innodata’s “AI Picks & Shovels” story about Innodata fundamentally relevant.
And one more important point: Hunterbrook has disclosed a long position in Innodata. The report should therefore be viewed as a well-researched thesis involving a conflict of interest, rather than as corporate confirmation.
For Tenbagger Capital, that’s exactly what’s interesting right now: Not just “Innodata is landing AI contracts,” but potentially Innodata is at the table with one of the largest AI investors of all time in the development of agent-based AI.
