Data Centers will be the Bulk of Nvidia’s Business in Five Years’ Time

NVIDIA seeks to address the world of connected devices on two fronts. First, it will serve the nebulous realm of the Internet of Things (IoT) with its numerous, low-value, sensor-based, data gathering devices, indirectly through the back-end. This will be accomplished by graphics processing unit (GPU)-based servers that will structure, process, and manipulate collected data to identify real-time trends and generate informed business decisions. Secondly, NVIDIA seeks to power the world of autonomous devices, the large and expensive endpoints that must perform mission-critical tasks regardless of whether they are permanently connected and being instructed or not. This client-side autonomous intelligence is provided directly using local GPUs that process a large quantity of data on the spot, knowing what to do then and there, supported whenever possible/appropriate by remote resources.

NVIDIA’s expansion beyond consumer gaming, into GPU products developed for client-side and server-based enterprise use cases, began in the automotive sector with self-driving cars. The company’s strategy is to target “lighthouse companies,” which are progressive and market-leading within their field, to develop vertical GPU-based services. These services will use parallel processing-powered, machine learning-driven artificial intelligence (AI) software to target the problems that enterprises seek to solve. NVIDIA intends to package and resell these services to lower-tier enterprises and will rely on strategic partners that act as integrators as its sales channel, with SAP cited as a key go-to-market partner. The inference engines that run AI algorithms require a data center and NVIDIA identifies the healthcare industry as having a huge need, making it one of the company’s next potential target verticals.

NVIDIA states that it has invested US$30 billion in GPU development. Graphics has always been a compute-intensive, high-volume application, with consumer gaming and professional graphics workstations being traditional markets for NVIDIA. The work done therein has subsidized the cost of GPUs for parallel processing in other industries, just as the ultra-competitive, high-volume smartphone market has subsidized the development and unit cost of modems and CPUs for portable devices. In both cases, bespoke development within vertical industries would have resulted in firms having to write off billions of dollars to engineer a similar competency. NVIDIA does not have to sell any inferencing chips to justify the existence of its parallel processing endpoint and server-side enterprise hardware, for which it has already paid.

Enterprise needs and the cloud opportunity have redefined NVIDIA’s mission statement: it is now an accelerated computing company. It has three stated focus areas of visualization (graphics), deep learning (AI), and high-performance computing.When an enterprise has a requirement for all three, there is an opportunity for NVIDIA. NVIDIA wishes to select a handful of verticals that are zero-size markets at present, of which it can take consequent ownership. It does not seek to play in markets that are already being served. However, this creates the risk of huge upfront investment in bespoke developments for speculative returns years down the line. NVIDIA’s business model is also being redefined, evolving from shipping units of chips to selling or leasing algorithms and/or big data processing capacity.CEO Jensen Huang states that, in 5 years, the “data center business will be biggest part of NVIDIA.”