Cognex pays about 7 times sales for the depth camera unit Intel spun out

An engineer who specced RealSense depth cameras into a machine last year has watched the supplier change hands twice in fourteen months. Cognex expects to close in the fourth quarter, subject to regulatory clearance ^[13]. What a team buys here is a part number and an SDK version pinned in a repository, and the assumption is that both will still be orderable at the next design freeze.

Retention and stock account for $106.5 million of the $606.5 million total, about 18 percent of what Cognex is handing over ^[2][3][1][2]. Most of the 180 employees hold shares as well and will collect at least $330,000 each from the sale, according to SiliconANGLE ^[5]. Spread evenly, the retention and stock work out near $592,000 a head ^[3].

RealSense says it expects $80 million to $90 million in sales this year, more than 50 percent above last year, and that it has been profitable for the past two quarters ^[8][9]. At the $85 million midpoint, the $600 million headline is about 7 times sales, and the $500 million of cash alone is about 5.9 times ^[4]. The same midpoint puts last year under $57 million ^[5].

Intel began the unit as an internal research project in 2014 and bought startups to build it into a force in 3D cameras and facial recognition, and it did not grow the way the chipmaker wanted ^[11]. It was spun out under Pat Gelsinger's restructuring, led by chief executive Nadav Orbach, with $50 million led by an unnamed private equity firm and participation from MediaTek Innovation Fund and Intel Capital ^[12]. Intel held on to 20 percent and one board seat ^[10]. On a pro rata split of the cash, that stake is worth roughly $100 million ^[6].

Cognex President and CEO Matt Moschner said his interest is the 3D depth perception and robot navigation, which he wants alongside the barcode and ID tag readers, defect inspection cameras and robot guidance systems Cognex already sells ^[17][18]. RealSense "brings a leading 3D perception platform with proprietary technology, a strong developer community and a compelling value proposition for customers," Moschner said ^[16]. The smaller facial authentication business is not part of the transaction and will be spun off separately ^[15].

The announcement gives no figure for RealSense's share of the depth camera market and no count of robots shipping with its cameras, so whether this consolidates a part that many robotics roadmaps depend on is still an open question. It does put a price on the revenue line and the developer community Moschner described ^[8][16]. For a team with these cameras designed in, the exposure turns on whether the sensor can be swapped without changing the mount and redoing calibration, and on how much of the perception stack sits on the vendor's SDK rather than the team's own code.

If you can second-source and you own your perception code, this is a footnote on a purchase order. If you can second-source but depend on the SDK, the thing to price is a version freeze. If the mount and calibration lock you to this camera and you own your code, the alternative is a mechanical redesign, so cost it now. If neither holds, the retention program sets your horizon: it runs three years from close, and the engineers it was written for vest at the end of it ^[3][13].