Overview
DeepMap built high-definition maps and positioning software for autonomous vehicles. Its maps represented road features in much greater detail than a typical navigation map. Its localization software helped a vehicle determine where it was within that map.[1]
NVIDIA acquired DeepMap in 2021 and incorporated its mapping work into NVIDIA DRIVE. The precise closing date and transaction price are not established in the sources below.[2]
Founding Story
James Wu and Mark Wheeler founded DeepMap in Palo Alto in 2016. Wu had worked on Google Earth, Apple Maps, and Baidu's self-driving platform. Wheeler had worked on positioning research, mapping imagery, lidar, and cloud mapping infrastructure. They saw the need for detailed maps that could support autonomous vehicles at scale.[3][4]
The founders' premise was that a self-driving system needs more than turn-by-turn directions. It must understand lane boundaries and other features, locate itself accurately relative to them, and receive updates as roads change. DeepMap focused on that infrastructure rather than manufacturing complete vehicles.[1]
Timeline
- 2016: Wu and Wheeler found DeepMap.[4]
- 2017: DeepMap announces a $25 million Series A led by Accel, with a16z and GSR participating. Total funding reaches $32 million.[1]
- 2018: TechCrunch reports a larger Series B and continued investment from a16z.[5]
- 2021: NVIDIA and DeepMap announce an agreement in June. In November NVIDIA says it acquired DeepMap earlier that year and describes the combined mapping system.[4][2]
What They Built
DeepMap described three components: detailed mapping, real-time localization, and infrastructure to serve maps at global scale. A map records features of a road. Localization compares a vehicle's observations with known map features to estimate its position and orientation. That position can help the driving system relate current sensor readings to lanes and road geometry ahead.[1][6]
Keeping a map useful is a separate challenge from creating it once. Road works, lanes, and signs change; a fleet must collect and process observations, identify changes, and distribute updates. NVIDIA later described combining DeepMap's survey mapping with fleet-sourced updates. That account describes the successor's approach, not proof that all promised coverage was achieved.[7]
Maps also have limits. A stored representation cannot replace live sensing of temporary obstacles or guarantee safe driving. Its usefulness depends on coverage, freshness, localization reliability, and how the rest of the vehicle system responds when information is uncertain.
Market Position
Target Customers
DeepMap sold mapping and localization services to organizations building autonomous vehicle systems. Its company profile named automakers, suppliers, trucking companies, and robotaxi businesses among target customers and listed several named customers. That profile does not disclose their contract values or deployment scale.[8]
Market Size
The sources reviewed do not establish a reliable market size for DeepMap's specific mapping services or a count of vehicles using them.
Competition
Other mapping companies and vehicle developers worked on the same problem; HERE and NVIDIA had announced mapping collaboration before the DeepMap deal. Some autonomous systems also place greater reliance on live sensing. NVIDIA's DRIVE platform still describes fleet-sourced mapping, so a new entrant faces an established successor as well as alternative approaches.[9][10]
Business Model
DeepMap described its mapping and localization offering as a service to vehicle developers. Its company materials do not establish prices, margins, renewal rates, or how the cost of surveying and updating roads compared with revenue.[8]
Traction
The company reported $32 million in total funding at its 2017 Series A and later raised a Series B. Its profile listed customers, but funding and named relationships do not establish commercial scale or profitability.[1][8]
Outcome and evolution
A mapping layer joined a broader vehicle platform
DeepMap's founders announced a proposed NVIDIA acquisition in June 2021. NVIDIA later confirmed that it had acquired the company and described integrating its expertise into DRIVE Mapping. A 2022 announcement named the combination of DeepMap survey mapping and fleet-sourced updates DRIVE Map.[4][2][7]
NVIDIA's current DRIVE software page still describes DRIVE Map, but the reviewed sources do not establish current coverage, deployment economics, or a one-to-one continuation of every DeepMap service. The sources do not reveal the relative value of the team, customers, and technology.[10]
Key Lessons
- Separate the map from localization. A detailed road model matters only if a vehicle can locate itself against it reliably.
- Budget for changes. A road map is a maintained service, not a one-time dataset.
- Build for the wider system. Mapping must work with sensing, planning, and safety behavior in the vehicle.
- Separate announced scope from delivered coverage. Road coverage and update economics require evidence beyond a product announcement.