Conceptual art: a warehouse mobile robot scans shelves and pallets with cyan and amber sensor beams; digital wireframes mark depth and velocity perception layers.
Physical AI often bottlenecks on sensing fidelity, not just model size., AI-generated cover

Alternate headlines

  1. Face ID veterans’ rematch: Lyte raises $165M at $1.6B for robot perception
  2. Robots need a trustworthy visual brain, not just bigger models: Lyte’s Series C
  3. From Kinect to Face ID to the factory floor: Lyte’s full-stack silicon bet

Lead

Lyte, the physical-AI perception company founded by former Apple Face ID / Microsoft Kinect engineers, closed a $165 million Series C led by Maverick Silicon at a $1.6 billion post-money valuation—about 3× the mark implied when it left stealth in January 2026. Total capital raised is about $272 million as LyteVision moves from prototypes into production deployments across inspection, logistics, and manufacturing.

What happened

Per the company release and TechFundingNews:

  • Round: $165M Series C led by Maverick Silicon; Fidelity (Series B lead), Atreides Management, Key1 Capital, Ora Global and others participated.
  • Valuation: $1.6B post-money; Lyte had already raised about $107M by its January 2026 stealth exit.
  • Board: Maverick Silicon managing partner Andrew Homan joins the board.
  • Product: LyteVision fuses a camera, inertial sensing, and a 4D sensor that measures distance and velocity into one synchronized stream at the silicon layer—aiming to avoid downstream sensor-stitching and calibration drift.
  • Use of proceeds: Scale sensing-silicon production, expand deployments, and hire across silicon, software, optics, and manufacturing.
  • Pedigree: Founders Alexander Shpunt, Arman Hajati, and Yuval Gerson are ex-Apple depth-sensing engineers; Shpunt previously co-founded PrimeSense, whose tech powered Kinect before Apple’s acquisition.
Physical AI has a sensing problem before it has a model problem. A robot cannot act safely on data that does not faithfully describe the world.
Alexander Shpunt, Lyte CEO (company release)

Why it matters

While frontier labs race on agents and gated cyber capabilities, Lyte is funding the perception substrate of physical AI. If warehouse robots still reconcile mismatched clocks and frames from off-the-shelf parts, even strong policy models will act on a drifted world state.

The competitive map is explicit: Rhoda AI bets on video-trained foundation models (reported ~$450M raise at ~$1.7B in March), while RealSense and Luxonis pursue more modular camera/depth stacks. Lyte’s wager is vertical integration from custom silicon to spatial software—native motion measurement instead of a downstream calibration tax.

Outlook

Watch the next two quarters for LyteVision’s field latency and miscalibration rates, silicon ramp, and whether full-stack ownership beats “models + modular hardware.” Physical-AI capital is plentiful; the referee is still fewer miscalibrated robots on the line.