NVIDIA's (NASDAQ:NVDA) data center segment is its fastest-growing segment and could be its largest segment within a few years, if its recent growth rate is any indication. After growing 145% to $830 million, or 12% of total revenue in the fiscal year that ended Jan. 29, the segment posted 186% year-over-year growth in the first quarter.

To further capitalize on the growing demand for deep-learning applications, NVIDIA recently unveiled its Metropolis deep-learning platform that will power artificial intelligence (AI) for various municipal services including parking and public safety as well as assist law enforcement. The Metropolis video analytics platform is what NVIDIA calls an "edge-to-cloud" solution, where the graphics specialist provides graphics processing units (GPUs) for "edge" devices like cameras that process information in real time and then send that information to supercomputers in the cloud.

A diagram showing a camera, server, and NVIDIA's DGX-1, Jetson, and Tesla GPUs and how they connect together to power AI cities.

NVIDIA Metropolis AI city computing solution. Image source: NVIDIA.

NVIDIA provides a complete solution

The complete end-to-end, or edge-to-cloud, solution is attractive because it enables GPU deep learning for all phases of the computing chain, from mobile ("edge") devices like cameras, drones, and robots, to the big workhorse computers in the data center.

Instead of a smart camera, powered by a central processing unit (CPU) and gathering basic information from the video feed and then sending a large batch of data through the cloud for processing on a data center, an AI camera is equipped with a GPU that allows it to process information locally in real time. A GPU-powered AI camera cuts the large amount of data down to a smaller amount, so by the time the data gets to the data center, there has already been some deep learning applied to it, which saves time. This also alleviates the problem of latency from sending a large batch of data wirelessly to another computer in the cloud, where information can get slowed down from a poor wireless signal, for example.

A side view of circuit board of NVIDIA's Jetson GPU.

NVIDIA Jetson GPU. Image source: NVIDIA. 

NVIDIA has a product for each stage of this process. Cameras equipped with NVIDIA's Jetson GPU filter the information in the video without human aid using AI. The metadata from the video is then sent to a server powered by NVIDIA's Tesla GPU for more advanced processing, which then can be sent to an NVIDIA DGX-1 supercomputer in the cloud for advanced deep-learning processing.

AI cities are a $2 billion opportunity

The advantage NVIDIA has in offering a complete, end-to-end solution has earned it more than 50 partners, among them market share leaders in video surveillance products, to create smarter cities with AI and deep learning. The number of cameras used by cities is expected to increase to about 1 billion by 2020, according to NVIDIA, and these cameras will capture an incomprehensible amount of data. NVIDIA pegs its AI city addressable market at about $2 billion by fiscal 2021. This opportunity is evenly split, with $1 billion for computer servers and $1 billion for cameras and sensors.

The end result of AI cities is greater efficiency of city management, which can also mean more efficient use of citizens' time. For example, AI sensors can let you know whether a parking space is available so you don't waste gas and valuable time. Other uses are faster facial recognition for law enforcement, better traffic management, pedestrian detection, and retail traffic analysis for businesses. AI cities could also optimize how energy is allocated to provide more efficient use of the power grid.

There seems to be a never-ending variety of ways for NVIDIA to apply its GPU computing technology to help solve the world's problems.

John Ballard owns shares of Nvidia. The Motley Fool owns shares of and recommends Nvidia. The Motley Fool has a disclosure policy.