Krishna RangasayeeFounder & CEO

We are in an amazing set of circumstances. While it looks like an enormous investment cycle today, we will look back in twenty or thirty years and realize this was just the very beginning of the AI journey.

We have seen massive AI adoption in the cloud, but the next step is moving compute closer to where we live. As AI touches and integrates with the physical world, it will expand its footprint and completely reshape our daily lifestyles.

Physical AI is entering mainstream focus and adoption. Our top three volume markets are robotics, automotive, and drones, though we are also seeing significant participation in medical, aerospace, and defense applications.

Our Gen 2 platform and silicon are now in production. We designed it to allow a friction-free transition from GPU AI infrastructure to our far more power-efficient and capable MLSoC platforms.

A SOM is a system-on-module board that helps customers speed up product development. We have launched a pin-to-pin replacement for the industry's leading GPU SOM, meaning developers can physically swap out a GPU for our module in seconds.

While we cannot disclose the full details of our partnership, memory is incredibly critical to scaling AI. Micron is an outstanding partner, and together we are focused on bringing combined value in compute and memory to our customers.

There is no one-size-fits-all solution for AI. As AI moves into reasoning and inference workloads, developers need heterogeneous compute rather than a GPU-only structure. Our platform integrates Arm processors, Synopsys DSPs, and our own proprietary machine learning accelerator.

Drones are transitioning from traditional first-person view units to fully AI-enabled, GPS-denied systems that can navigate without knowing the environment beforehand. Due to current geopolitics, we expect massive demand for AI drones and anti-drone technology.

While we participate in the humanoid space, my view is that the adoption of humanoids will happen at a much slower pace than the market currently expects. However, the AI-enablement of traditional mobile, warehouse, and logistics robots is scaling rapidly today.

We are seeing traditional tools like lawnmowers become autonomous and agentic. With our technology, users can speak directly to a mower to set schedules, and the machine itself can predict when its rotors need replacement, removing the need for manual monitoring.

Unlike enterprise software applications where companies are searching for revenue and product-market fit, edge AI targets existing, known markets. We are introducing new productivity and capabilities to established industries, so there is no market risk.

Palette Ne acts as an agentic wrapper on our software development kit. It allows developers to build complex edge systems in plain English rather than complex code, reducing the time to migrate and bring applications to production from months to hours.

GPUs are graphics cards built primarily for data centers. They are far too power-hungry and inefficient for edge applications. Physical AI requires ruggedized, cost-effective, and power-efficient silicon designed specifically for industrial edge environments.