Artificial intelligence gradually learns to act on the physical world. But between a model capable of making a decision and a machine capable of executing it, an essential technological layer remains largely absent.
Autonomous drones, industrial robots, driverless vehicles or robotic medical systems must constantly transmit video streams, sensor data and commands with delays of a few milliseconds. A constraint that current Internet infrastructures, designed for communication or content distribution uses, have never been optimized to manage.
As the artificial intelligence industry moves toward robotics, autonomous systems and defense applications, this issue is becoming a strategic issue. After models, semiconductors and data centers, a new category of players is now trying to build the infrastructure that will allow AI to interact with the real world.
It is on this technological layer that Kyber, a company founded by Jean-Baptiste Kempf, creator of the VLC multimedia player and main architect of FFmpeg, one of the most used software libraries in the world for video processing, is positioned. The company announces a Seed fundraising of five million dollars, or approximately 4.3 million euros, led by LIGHTSPEED VENTURE PARTNERS with the participation of OVNI CAPITAL and KIMA VENTURES.
Behind this operation lies an ambition which goes far beyond the framework of robotics software. Kyber intends to develop the real-time infrastructure allowing human operators and autonomous agents to control, observe and train machines from any point on the globe.
An infrastructure designed for real time
Most existing systems today rely on technologies developed for the consumer Internet, including WebRTC, a widely used protocol for video communications.
These tools were designed to allow humans to communicate and were not designed to permanently coordinate robots, drones or autonomous systems operating in critical environments.
In the physical world, a few hundred milliseconds can represent several meters traveled by a drone or autonomous vehicle before a command is executed. In certain industrial, medical or military applications, this delay becomes incompatible with operational requirements.
Kyber claims to be able to reduce this latency to around eight milliseconds by synchronizing all the flows necessary for the operation of an autonomous machine within the same infrastructure.
The objective is to provide developers with a single software layer to simultaneously manage video, audio, sensor data and control commands, while ensuring temporal coherence between these different streams.
The same logic as VLC and FFmpeg
Jean-Baptiste Kempf’s ambition is a continuation of the projects that have made his reputation. VLC and FFmpeg have not established themselves as mainstream products in the traditional sense of the term. These technologies have become reference infrastructures, integrated into millions of applications, services and equipment around the world.
Their distribution was based on open source, which favored their adoption by developers and businesses. Over time, these software building blocks have established themselves as de facto standards in the global video ecosystem.
The chosen model is based on a so-called “open core” approach: part of the technology is distributed as open source in order to encourage its adoption by the ecosystem, while the advanced functionalities, administration tools, support and commercial licenses constitute the basis for monetization with companies.
For investors, this strategy is of particular interest. In the software industry, the technical standards that are imposed at the start of a technology cycle often capture a significant portion of the value created by the entire ecosystem. The question is then no longer how many robots or drones will be sold, but how many will tomorrow use the software layer that connects them to their operators, their sensors and their artificial intelligence systems.
A thesis that appeals to AI investors
The support of Lightspeed illustrates this reading, the American fund has increased investments in artificial intelligence infrastructures, business software and technologies linked to autonomy. For Antoine Moyroud, partner at Lightspeed Venture Partners, the observation is simple: models are progressing rapidly, but the infrastructures allowing them to connect to the physical world are still significantly behind.
This analysis joins a broader trend observed in the industry. After funding the models, GPUs, and data centers, investors are starting to look at the software layers that will enable autonomous systems to operate in real-world environments.
Robotics, defense, industry and telecommunications are today the primary target markets.
Infrastructure for the machine economy
The rise of autonomous systems could profoundly change the very nature of digital infrastructures.
The Internet of the last two decades was primarily designed to connect people, content and applications. The economy that is now emerging will need to connect millions of machines capable of observing their environment, making decisions and acting autonomously.
In this context, communication protocols, real-time infrastructures and control layers could become as strategic as cloud platforms have been for the software economy.
It is on this hypothesis that Kyber’s bet is based today. Having helped build some of the fundamental building blocks of digital video, Jean-Baptiste Kempf is now looking to develop one of the infrastructures that could support the next phase of artificial intelligence and create the successor to WebRTC for the era of robots, drones and physical AI.
