VLC Creator Raises $5M for Robot Control Startup Kyber

The technical lineage is easy to trace. Kyber began as a side project Kempf built while he was CTO at the cloud gaming startup Shadow, and its approach to killing lag comes straight out of video-streaming engineering. But the other half of the work is IoT optimization — tuning performance to whatever compute a given device has on board, and doing it across a huge number of machines at once.

Why scale is the real bet

Kempf is open that the technology is not unique on its own. Companies with deep enough pockets have already built remote-control software for their specific needs, remote driving being the obvious example. His argument is about volume.

The largest fleets today, he says, have maybe two or three thousand vehicles. Kyber is being built for a future where an operator might manage millions of them. That is a different engineering problem, not just a bigger version of the same one.

That jump in scale changes what matters most. Observability — knowing for certain that systems are actually working — becomes far more important once AI agents, rather than people, are running entire fleets and networks. Even at small scale there is an immediate payoff Kempf points to: not having to physically walk up to every device just to push a software update.

As fleets grow from a few thousand machines to millions managed by autonomous agents, the ability to confirm that every remote device is truly online and responding stops being a convenience and becomes a safety requirement.

The open-source playbook, again

True to the way Kempf built VLC, Kyber's core project is open source. The company makes money by selling a productized, enterprise version on top of it. That structure means the user base will almost certainly be far larger than the paying-customer base — a lot of people will run the free core without ever cutting a check.

It is not only software, either. Like Palantir, Kyber offers hands-on custom deployment through forward-deployed engineers, often shortened to FDEs. Those engineers already make up a large share of the team, which currently stands at 25 full-time staff. The company is headquartered in Paris but keeps offices in San Francisco and Singapore to chase a global client base.

Where Kyber is focusing first

Kyber says it is already in commercial deployment with customers across defense, telco, robotics, and AI. To keep its small team focused, it has narrowed its attention to three segments.

• Robotics — the headline use case, riding the physical-AI wave.
• Drones of every kind — aerial and otherwise.
• Remote IT access — the least glamorous segment, but the one where Kempf says demand has been strongest.

That third bucket is worth dwelling on. Kempf says Kyber wants to be more than a Citrix challenger, but even that comparison points at a large addressable market. The careers page sums up the pitch bluntly: the companies that tried to solve remote access spent years and tens of millions building custom tools they will never share, and Kyber wants to build the version everyone else can use.

The numbers at a glance

What happens next over the coming days

With the round public, expect Kyber to lean into hiring, especially more forward-deployed engineers, since that group already anchors the team and the custom-deployment model depends on it. A 25-person company spread across three continents will feel pressure to staff up fast in San Francisco and Singapore to service enterprise accounts in different time zones.

Watch for early customer names to surface. The company says it is already deployed in defense, telco, robotics, and AI, but specifics have stayed vague, and named deployments are usually the next thing a freshly funded startup discloses. The defense angle in particular tends to draw scrutiny, so any concrete contract there will be closely read.

The broader signal is what Lightspeed is really buying. If hundreds of millions of robots and drones do end up roaming streets, the plumbing that controls them at scale becomes valuable infrastructure. Kyber is betting that the unglamorous layer underneath physical AI — the part that keeps latency low and confirms machines are actually responding — is where a durable business sits. The next few funding announcements in this space will show whether other investors agree.

Source: TechCrunch