Looking back at NATIX last year and our achievement of spreading the word of DePIN through Drive&, we realized how far we ventured into unknown waters. The concept of a Decentralized Data Curation Network was new, and its potential was overwhelming. Drive& was a humble beginning — a chance to prove that scalability through Decentralized Infrastructure Networks (DePIN) and existing hardware can catch the attention of mapping giants. It was a huge success. Not only does Drive& have over 260K(!) users, but it also puts NATIX on the map (pun intended) as the biggest mapping DePIN network globally, according to Messari’s “State of DePIN 2024” report.
We thought: “Why should we stop there? The potential is there, so why not use it to tackle bigger problems?”
So, how do we leverage NATIX’s Internet of Cameras and our DePIN scalability via a commodity hardware approach to bring tangible change? To answer that question, we need to circle back to January, when NVIDIA's CEO Jensen Huang planted the term “Physical AI” in everyone’s radar at CES 2025.
When OpenAI’s ChatGPT was first introduced to the world, it did so by storm. Some started fearing for their jobs, others warned that a Terminator-like future is inevitable, and a few wasted no time in testing its limits and declared that AI will never be able to replace us. Nevertheless, generative AI tools are now transforming entire industries, but, as NVIDIA put it, they’re limited in their grasp of the physical world and its rules.
That’s when NVIDIA's CEO Jensen Huang stated that tech giants are shifting from Agentic AI to Physical AI, meaning autonomous machines that are able to perform tasks in the real world. However, achieving large-scale, widespread Physical AI is easier said than done. For an AI to be ready for the real world takes time, computing power, and an incredible amount of data.
When it comes to generative AI, the process is quite straightforward — with over 402 million terabytes generated every day on average, harnessing the necessary data is fairly simple. When it comes to Physical AI, the process is more complicated.
Physical AI, unlike generative AI, must be able to perceive, process, and navigate in the real world. For humans, that process happens instinctively — what our eyes see is instantly processed in our brain, and we can act upon it. For Physical AI models to act accordingly is far more complicated. Two main problems are giving the Physical AI world a massive headache: compute power and data.
The process of preparing an AI model for the real world first requires training the AI. For that, we need to provide it with data and let it run for as many times as needed until it learns. Not to mention that training an AI is not a one-time thing. You train, then test/validate, and back into the training grounds you go to refine the results including scenario generation. This process consumes an incredible amount of data and computing power. With the recent rapid acceleration of AI, this has led to an unprecedented demand for high-performance computing resources (mainly high electrical consumption and GPUs), resulting in a severe GPU shortage, and without access to sufficient GPU power, innovation slows, and costs skyrocket.
The data aspect of training Physical AI models is no less problematic. Even NVIDIA, the most notable promoter of Physical AI innovation, has identified data scarcity as a key bottleneck for the development of physical AI. To interact with the physical world, AI requires visual data, which is extremely difficult to collect, resulting in a serious lack of data, giving companies promoting Physical AI a real headache.
Attempting to solve the problems arising from a lack of computing power is indeed problematic. Companies are fighting for GPUs, and while having NVIDIA on the cap table always helps, projects are scrambling to find alternatives to the high electrical consumption of AI training. Cloud providers are already leading the charge in tackling the shortage of GPUs, but a tangible solution is yet to be found. On the other hand, the scarcity of data has a remedy.
Decentralized Physical Infrastructe Networks (DePIN) crowdsource data from its network of participants, rewarding users and opening up access to the data collected. A sub-category within DePIN — Data Curation Networks (DCN) captures and curates data directly from users, offering a solution for the vastly growing demand for data in the AI market, particularly Physical AI, where data is paramount for models to interact with the real world successfully.
NATIX Drive& is proof that DePIN is a viable alternative to traditional data collection methods. With over 260K drivers contributing geospatial data, the value NATIX offers to the mapping industry is unmatched; however, Drive& is an alternative, albeit superior, to already existing geospatial data collection solutions. As mentioned before, there is no “traditional” solution to data collection for Physical AI, which is why it is considered one of the industry’s biggest problems.
Identifying the dire need for data that Physical AI innovators can use was instant. The need for data existed long before the term Physical AI was born, with autonomous driving companies relying on building expensive vehicles to collect real-world driving footage, and that was only for the training phase of the AI. Then came the race for Physical AI in January 2025, and everyone tried to capitalize on the opportunity, but there were no answers in sight. Even Learning to Drive (L2D), the world’s largest open-sourced dataset, was released with approximately 5K hours of driving collected only in Germany, but that took three years to collect, providing no real solution. Enter VX360.We know our way around cameras, the king of sensors, and Physical AI needs to be able to see. Once again, relying on our principle of scalability by leveraging commodity hardware, we realized that the opportunity to solve the unsolvable was right in front of us. Tesla’s widespread availability as the world’s best-selling car for two consecutive years, along with its ability to capture 360° imagery via its set of cameras, provided the perfect infrastructure. Build a device that can tap into that footage, sprinkle some DePIN to reward our userbase, and your data collection device that can advance Physical AI innovation is complete. VX360 is revolutionary.Drive& is important for the mapmaking industry, whereas VX360 is critical for Physical AI. The value of the data generated by our smartphone drivers is immense, but the value of the data captured by our network of Tesla drivers is incomparable. It is no coincidence that shortly after VX360 was released, the term Decentralized Physical AI (DePAI) emerged, a new term that brings together decentralized projects aimed at powering Physical AI innovation, and NATIX is a pioneer in that space spearheading the rise of DePAI.
With Physical AI companies facing some major headaches, NATIX positions itself as the remedy and a leader in the DePAI space. Not only is VX360 one-of-a-kind in its ability to solve the unsolvable (with traditional data collection methods), but the revenue generated by the device presents a win-win scenario for NATIX users. Not only do our users earn for contributing data, but through our Protocol Revenue, a major chunk of our earnings is used to sustain the ecosystem.DePAI ensures Physical AI has a future in the real world, and NATIX will be a leading force in its advancement. Where there is potential for innovation is where we thrive because DePIN is the frontier.
