Lyte Raises $165M to Scale Physical AI Perception

A robot can have a sophisticated AI model and still fail when its sensors disagree about where the world is or how quickly it is moving. Lyte has raised a $165M Series C to turn that integration problem into a shared perception platform for robotics. Maverick Silicon led the round, which values the Sunnyvale company at $1.6B post-money and brings total disclosed funding to $272M.

Lyte combines custom sensing silicon, 4D coherent vision, high-resolution imaging, inertial sensing, and spatial software on one synchronized timeline. The company says LyteVision has entered production and is shipping to customers across inspection, logistics, and manufacturing. Those customers remain unnamed, but the transition from stealth to production gives the financing a clear operating purpose: scale the systems that make perception reliable enough for robots to move beyond controlled demonstrations.

The September 2, 2026 financing also adds Maverick Silicon managing partner Andrew Homan to Lyte's board. Fidelity Management & Research Company, which led the previous Series B, participated alongside Atreides Management, Key1 Capital, Ora Global, and other existing and new investors.

What Lyte raised and who backed it

Lyte announced the $165M Series C at a $1.6B post-money valuation. The round more than doubles the $107M in aggregate Series A and B financing that Lyte disclosed when it emerged from stealth in January 2026. Primary sources do not break the earlier total into separate round amounts, so the clean capital accounting is $272M raised since the company's 2021 founding.

Maverick Silicon led the round. Fidelity returned after leading Lyte's Series B, while Atreides Management, Key1 Capital, and Ora Global joined the named syndicate. The financing will be used to expand production of Lyte's sensing silicon and LyteVision platform, advance AI and perception capabilities, broaden deployments, and hire across silicon, software, optics, manufacturing, operations, go-to-market, and senior leadership.

Homan's board addition gives the lead investor a formal role as Lyte moves from early production into a heavier manufacturing and commercialization phase. Chairman Avigdor Willenz brings a separate history around foundational semiconductor businesses, including Galileo Technology, Annapurna Labs, Habana Labs, and Astera Labs.

Why physical AI begins with perception

Robots do not receive the physical world as one clean stream. Cameras capture color and texture. Depth systems measure distance. Inertial sensors track movement. Each component can be accurate on its own while the complete system still struggles with timing, calibration, occlusion, vibration, weather, or changing light. The application team has to turn those separate signals into a coherent scene quickly enough for a machine to act.

That burden matters because physical-AI errors have consequences that software demonstrations can hide. A delayed answer in a chat window is inconvenient. A delayed perception update near a person, pallet, vehicle, or robotic arm can become an operating and safety event. Lyte's market thesis is that the model problem arrives after the sensing problem: a robot cannot reason reliably from a fragmented account of the world.

LyteVision addresses that handoff by integrating 4D coherent vision, RGB imaging, inertial sensing, custom silicon, and spatial software. Lyte says the platform fuses those inputs at the hardware level onto one synchronized timeline. It measures position and motion natively rather than asking downstream software to reconstruct every relationship after capture.

The founders behind Lyte

Lyte's three co-founders have spent much of their careers moving perception technology from research into mass-market systems. Alexander Shpunt, CEO, co-founded PrimeSense and served as its CTO. PrimeSense's 3D sensing powered Microsoft Kinect before Apple acquired the company in 2013. Shpunt later held senior advanced-technology roles at Apple.

Arman Hajati, Lyte's CTO, previously led architecture work for Apple's Taptic Engine and incubation efforts inside Apple's Advanced Technology Group. Yuval Gerson, VP Engineering, held senior R&D and architecture roles at Apple, PrimeSense, Rafael, and several startups. Lyte describes their earlier work as foundational to Face ID and says the underlying perception technology reached more than 1B people.

That history is relevant because robotics companies face a problem the consumer-electronics industry already learned to respect: sophisticated sensing becomes valuable only when it can be manufactured, calibrated, integrated, and trusted at scale. Lyte is taking the founders' experience with highly integrated perception systems into machines that operate around factories, warehouses, loading docks, and other dynamic environments.

From synchronized sensing to production robotics

Lyte's platform is designed to give robotics teams a more complete starting point. Instead of assembling sensors, synchronization logic, processing, and perception software as separate projects, a customer can use a stack designed from the transistor through the spatial-data layer. The customer is buying engineering time as much as hardware. A more integrated perception system can let teams concentrate on robot behavior, industrial workflows, and deployment economics rather than rebuilding sensor fusion.

The company has also connected LyteVision to NVIDIA's physical-AI ecosystem. In June, Lyte said it joined NVIDIA Halos for Robotics and the NVIDIA Halos AI Systems Inspection Lab. A separate collaboration with NVIDIA Holoscan Sensor Bridge sends high-bandwidth perception data directly into GPU memory for digital twins, simulation, training, and inference. These relationships do not independently prove the platform's commercial scale, but they place Lyte inside the tooling and safety architecture used by robotics developers.

Lyte reports that it is shipping to leading customers across inspection, logistics, and manufacturing. The company has not identified those customers or disclosed shipment volume, revenue, margins, or deployment-level performance. That evidence boundary matters because moving from a handful of production programs to reusable infrastructure across many robot categories requires manufacturing consistency, support, safety validation, and repeatable integration.

What the $165M changes

The Series C gives Lyte enough capital to make production capacity part of the product strategy. Custom silicon and optics demand longer development cycles, supply-chain planning, packaging, testing, and customer support. Lyte expects to expand a team reported by Bloomberg-derived coverage at about 120 people, with hiring across technical, operational, commercial, and leadership roles.

For investors, the opportunity is larger than a single robot design. If perception remains fragmented, every robotics company pays the same integration tax before its application can scale. A trusted, reusable layer could serve humanoids, mobile robots, industrial systems, inspection platforms, and other autonomous machines without Lyte needing to build those machines itself.

That is also the obligation created by the valuation. The $1.6B post-money figure prices in more than strong engineering history. Lyte now has to show that its integrated stack can travel across customers and environments while preserving timing, safety, and reliability. Every production floor will introduce a different mix of motion, clutter, light, vibration, and human behavior.

The physical-AI market will keep celebrating models that reason and robots that move. Lyte's next chapter will unfold one layer earlier, where photons, velocity, and timestamps have to become a version of the room that a machine can trust before the machine decides what to do.