Ex-Apple Face ID engineers raised $165 million to give robots a trustworthy sense of the world
Lyte's Series C, led by Maverick Silicon, triples the perception-hardware startup's valuation to $1.6 billion -- its bet is that every category of physical-AI robot will need custom sensors and silicon that make 'is this real' a solved problem before a robot ever acts on it.
Lyte, a robot-perception hardware startup founded by former Apple engineers, announced a $165 million Series C on September 2, 2026, led by Maverick Silicon and bringing its post-money valuation to $1.6 billion -- more than triple its prior mark. Fidelity Management and Research, which led Lyte's Series B, returned alongside Atreides Management, Key1 Capital, and Ora Global, taking the company's total capital raised since its 2021 founding to $272 million.
The pedigree behind the pitch
Founders Alexander Shpunt, Arman Hajati, and Yuval Gerson worked on advanced sensing and perception technology at Apple; Shpunt previously co-founded PrimeSense, whose 3D-sensing technology powered the original Microsoft Kinect before Apple acquired the company in 2013 and folded that work into what became Face ID. Lyte builds custom silicon, multimodal sensors, and spatial software that let a robot determine where it is and what's moving around it -- CEO Shpunt's own framing: "physical AI will create entirely new categories of robots, and every one of them" will need trustworthy perception to act on.
Why perception is the boring, load-bearing part
It's easy for robotics funding coverage to gravitate toward the flashiest capability -- a humanoid folding laundry, an arm performing surgery -- and skip past the sensing layer that has to be right before any of that is safe to run unsupervised. Lyte's bet is essentially that the more autonomous a robot's decision-making gets, the more its perception hardware needs to be treated as a distinct, auditable component rather than an assumed-solved input, which is the same logic this site keeps applying to software agents: a system's competence at the task in front of it says nothing about whether the sensing or access layer underneath is actually trustworthy, and that layer is exactly where a well-funded, narrowly focused vendor can do real work other AI headlines skip past.