About MCT
Native Infrastructure for the Physical AI Era
MCT is committed to becoming native infrastructure for the Physical AI era. Guided by a "data-driven, hardware-software integration" full-stack strategy, MCT has built a complete technology chain spanning perception, computation, data, simulation, models, and execution. With over 1 million vehicles deployed, more than 5 billion kilometers of real-world mileage validated, and deep co-creation with leading automakers, MCT has established a "chip–hardware–model" vertical closed-loop with a significant time-to-market advantage. Just as NVIDIA defined computing infrastructure for the digital AI era, MCT is taking this large-scale, production-proven methodology as its cornerstone to systematically enable trillion-dollar sectors including embodied AI, low-altitude economy, and industrial intelligence.
The physical world is not a projection of the digital one
Language models learn symbolic logic. The physical world runs on energy, mass, and causality. This is not a difference of degree — it is a difference of starting point.
Every digital AI system operating today — writing text, generating images, holding conversations — works within symbolic space. When AI steps off the screen and into the physical world, it needs a fundamentally different foundation.
The physical world has no 'close enough.' A robot entering a factory that is off by a centimeter causes an accident. An eVTOL flying over a city that loses positioning is a disaster. Infrastructure designed for digital AI has been misaligned with the physical world from day one.
“This trillion-dollar sector doesn't even have its own native infrastructure yet.”
Three Dimensions of Physical AI-Native Infrastructure
Five core capabilities enabling intelligent systems to run, evolve, and scale in the real world
Physical AI-native infrastructure is built on three mutually reinforcing dimensions:
Perception, computation, and control-execution form the real-time operating system; data and simulation form the continuous evolution system; reliability and safety provide engineering assurance across the full stack.
Real-Time Operating System
Enabling intelligent systems to perceive the real world, make real-time decisions, and drive physical action
Perception
Accurate perception of the real world is the starting point for all intelligent decisions.
Multimodal signals — visual, tactile, attitude, position, and force — must be continuously fused under a unified time and spatial reference, forming a stable, continuous, and trustworthy environmental input.
Native Compute
Physical interaction demands a native compute architecture.
Physical AI must complete multi-sensor fusion, state estimation, real-time inference, and action decisions at the edge, while maintaining low latency, low power, and high reliability.
Control & Actuation
Intelligence only enters the real world when it translates into stable, controllable action.
The system must convert perception and decision outputs into precise, continuous physical behavior, maintaining real-time closed-loop control under varying loads, environmental disturbances, and structural errors.
Continuous Evolution System
Letting real-world data and simulation capabilities drive ongoing model and system iteration
Data & Simulation
Real data calibrates boundaries; simulation accelerates evolution.
Real-world data continuously flows back for training, calibration, and validation. Simulation expands training scale and covers complex and long-tail scenarios — together forming a closed-loop iteration cycle for data, models, and systems.
Engineering Assurance Framework
Keeping the system reliable, controllable, and deliverable in complex real-world environments
Reliability & Safety
Reliability and safety determine whether Physical AI can scale.
From sensors, chips, and models to control and actuation, the system must have redundancy design, fault detection, graceful degradation, functional safety, and mass production validation capabilities — providing engineering assurance across the full stack.
MCT's Three Core Advantages
“Our customers don't buy our chips — they buy the years of hard lessons we walked through with them.”
1M+ vehicles, 5B+ km — built alongside leading customers, kilometer by kilometer. Those kilometers cannot be compressed, and the time cannot be replicated. The data assets and engineering knowledge accumulated here can only be earned by walking the same road again.
MCT has treated chips, hardware, and models as one unified problem from day one. This is not just a strategic choice — it is an organizational starting point. The structural barriers to vertical integration inside large companies cannot be solved with resources alone.
From rule-based to end-to-end, from end-to-end to VLA and world models — MCT has lived through every paradigm shift in autonomous driving. Each evolution happened inside real mass production projects. This systematic knowledge cannot be bought.
The standardization window for Physical AI infrastructure is opening — first movers define the rules.
MCT is committed to becomingPhysical AI-native infrastructure
Starting from real-world data, we continuously refine the technology chain — from perception, computation, data, simulation and models to execution — driving Physical AI from 'capable' to 'reliable.'
·Accumulate real-world data
Continuously building data capabilities across hand motion, whole-body posture, vision, positioning, and force/haptic feedback from the real physical world.
·Validate native chip direction
Advancing R&D on perception and compute chips, validating multi-sensor fusion, high-precision synchronization, low power consumption, and real-time processing.
·Build platform capabilities
Through co-creation with lead customers and validation across representative scenarios, establishing standardized products, data services, and simulation platform capabilities.
MOJANDA: MCT's proprietary automotive-grade GNSS chip, the origin of positioning capability
The MOJANDA series is the positioning compute core of all MCT products, and the starting point of the "chip–hardware–model" vertical integration strategy. Built on a deeply customized RISC-V architecture, it delivers ~30% better positioning accuracy than international leaders in heavy-occlusion scenarios — from project launch to mass production in just 21 months.
Wujin Hi-Tech District, Changzhou, Jiangsu
Annual capacity: 1M+ automotive-grade units