Physical AI startup Mowito raises $3M to scale foundation models for robot arms

As large language models reshape software, Mowito is applying the same foundation-model philosophy to hardware. The startup's approach—training AI on diverse assembly demonstrations to achieve few-shot task learning on robot arms—represents a key frontier where AI meets the physical world, and its $3M pre-seed round signals growing investor conviction that Physical AI is the next major platform shift.

In a significant move for the industrial robotics sector, Bengaluru- and Detroit-based startup Mowito has raised $3 million in pre-seed funding to scale its AI foundation models for robot arms. The round, led by Version One Ventures with participation from All In Capital, Unisol, iSeed, and prominent angel investors including PyTorch co-creator Soumith Chintala, underscores growing venture confidence in 'Physical AI'—the application of large-scale learning models to tangible, real-world tasks. Founded in 2024, Mowito addresses a long-standing pain point: the rigidity of industrial robots, which demand costly, time-consuming reprogramming for every product or process change. By enabling robots to learn tasks simply by observing human demonstrations, Mowito aims to eliminate the software bottleneck that manufacturer after manufacturer has cited as the primary barrier to automation agility.

“

In a significant move for the industrial robotics sector, Bengaluru- and Detroit-based startup Mowito has raised $3 million in pre-seed funding to scale its AI foundation models for robot arms.

The timing is apt. The global industrial robot market, valued north of $50 billion, is expanding but faces a skilled-programming shortage and escalating demand for high-mix, low-volume production. Automakers and electronics contractors, in particular, are pivoting to rapidly shifting product lines, where traditional robot teaching methods can take days or weeks and require specialized engineers. Mowito's foundation models, trained on a diverse corpus of assembly and manipulation tasks, can be retasked in minutes, slashing changeover downtime and labor costs. This approach mirrors the zero-shot and few-shot learning breakthroughs seen in language and vision models, but applied to kinematic chains and precision thresholds of ±0.1mm or less—a significantly harder problem given the physical constraints and safety imperatives.

The startup's early traction is noteworthy. Mowito disclosed that its AI-powered robots are already operating on manufacturing lines at a Fortune 500 automotive company and one of the world's largest electronics contract manufacturers. These deployments involve high-precision assembly applications, serving as powerful proof points for the technology's robustness in mission-critical environments. The dual presence in Bengaluru (R&D hub) and Detroit (the automotive heartland) positions Mowito to both innovate cost-effectively and maintain close customer intimacy with North American manufacturers.

The fresh capital will be channeled into three areas: accelerating US expansion to capture the large, automation-hungry American market; strengthening engineering and go-to-market teams to accelerate product development and sales cycles; and scaling deployments across automotive and electronics manufacturers, which together represent a massive addressable market. For supply chain operators, this means a future where robot cells can be re-purposed for new products without a six-figure integration project. For investors, Mowito sits at the confluence of deep tech, SaaS-like recurring revenue models, and manufacturing's digital transformation.