Nvidia Corp. has inked a deal to acquire DeepMap Inc., a well-funded startup with mapping technology that helps autonomous vehicles navigate the road more reliably.

Though the terms of the deal were not disclosed in the Thursday announcement, it’s safe to assume Nvidia is paying a fairly substantial sum for the startup. DeepMap has raised more than $90 million in funding since launching six years ago from investors that included Andreessen Horowitz.

The San Francisco-based startup makes software that automakers can use to create high-fidelity maps of roads. Such maps are essential for autonomous vehicles, which rely on them to plan the best way to reach a destination and identify potential obstacles. The more data a vehicle’s artificial intelligence driving system has about roads, the more reliably it can navigate them.

DeepMap has announced two mapping-related offerings since October, which Nvidia is set to absorb as part of the deal. 

The first offering is called DeepMap HDR. According to the startup, DeepMap HDR can help automakers generate maps and continuously update them over time to ensure autonomous vehicles have an accurate view of their environment. More recently, the startup introduced an offering called RoadMemory that can automatically build maps using the sensors attached to vehicles.

DeepMap’s technology will expand Nvidia’s growing lineup of products for the auto sector. The company has built an end-to-end set of solutions for autonomous vehicles that includes not just chips but software tools for building, training and testing AI driving systems.

The centerpiece of Nvidia’s auto portfolio is its Drive AGX series of system-on-chip processors. They can be deployed inside autonomous vehicles to power the onboard AI software that makes navigation decisions. The newest chip in the Drive AGX series is the Orin, which was described as seven times faster than its predecessor when it debuted. Each Orin chip combines graphic processing elements with Arm Ltd. central processing units to deliver maximum performance of 200 trillion operations per second.

Before an AI driving system is ready to be deployed in vehicles, an automaker has to run the system through simulated road tests to ensure it’s capable of operating reliably. For that aspect of development projects, Nvidia offers its Drive Sim software platform. The platform enables automakers to put their AI algorithms to the test on simulated roads that they can parallelize across multiple graphics processing units to speed up experiments.

Drive Sim draws on Nvidia’s ray tracing technology, which the chipmaker originally developed for its consumer graphics cards. Ray tracing is a rendering technique popularized in the film industry that provides the ability to generate highly realistic light and shadow effects. Drive Sim uses it to increase the realism of AI simulations by adding in details that would be difficult to render otherwise, such as dim shadows inside a tunnel or light rays reflected off a truck.

The other components of Nvidia’s autonomous vehicle portfolio cover subsequent phases of the development process. When it’s time to build a prototype vehicle to test an AI driving system, automakers can save time by buying chipmaker’s Drive Hyperion kit, which offers a pre-packaged suite of vehicle sensors. To increase the underlying AI algorithms’ accuracy, automakers can train them on Nvidia’s DGX machine learning appliances.

The mapping technology Nvidia is set to gain through the acquisition of DeepMap will improve its ability to address another key project requirement of companies building autonomous vehicles. At the same time, the deal could help the chipmaker step up competition against Intel Corp.’s Mobileye unit.

Mobileye sells chips and other technology for building partially autonomous driver assistance systems that automate tasks such as parking. It’s also working toward assembling a technology stack for building fully autonomous vehicles. As part of the effort, the Intel unit  has created a mapping service that generates maps of roads using sensor data collected by vehicles from several major automakers.

With DeepMap’s mapping technology, Nvidia can more directly take on that part of Mobileye’s business. The technology could in the process increase the overall value of its end-to-end product lineup for autonomous vehicles, thereby potentially helping it sell more AGX Drive systems-on-chip in the long run.

Nvidia expects to close the acquisition in the third quarter. The startup’s technology will extend the company’s existing Drive Mapping offering, which allows autonomous vehicles to assemble a map of a road from data gathered by their onboard sensors. 

Image: Nvidia