Announced positioning, filed reality. Luminar is sold to the market as a LiDAR hardware story, and hardware trades at hardware multiples. But Luminar Technologies' March 31, 2020 grant US10606270B2 claims something the spec sheet does not advertise: a method for controlling an autonomous vehicle using cost maps built from perception data.
Follow the margin, not the cap table press release. Cost maps are the software layer that translates a point cloud into a steering decision. The CPC tags — G05D 1/0088, G01C 21/3492 route costing, G06N 3/08 neural inference, plus a stack of G05D path-planning classes — describe planning software, not a laser.
The reconciliation matters because a sensor company that owns the planning software has a different gross-margin profile than one that ships glass and diodes. Software attach is what turns a commoditizing hardware line into a defensible platform. This patent is Luminar putting that ambition on the public record in 2020, years before its revenue mix could prove it out.
The skeptic's check: filing a planning patent is not the same as booking software revenue. There is no line in a sensor maker's filings that says "cost-map licensing." The patent tells you intent and direction, not dollars.
Still, for a deal desk valuing the autonomy supply chain, the question this grant forces is the right one: are you paying for a sensor, or for the software stack that makes the sensor worth integrating? Luminar clearly wants the answer to be the latter.
The takeaway is that the most interesting autonomy IP often sits one layer above what a company is famous for. Reconcile the valuation to where the patents point, not to the product on the booth.