NVIDIA Doubles Down on Industry-Specific Solutions to Accelerate Enterprise Artificial Intelligence Deployment

Traditionally, Artificial Intelligence (AI) technological advancements were propelled by two very distinct camps of market players. On the one side, software developers would push the boundary by introducing cutting-edge machine-learning and deep-learning models, techniques, and frameworks, such as Open AI and Meta. On the other side, semiconductor companies would introduce new processors that can support the ever-growing demands for AI training and inference workloads, leading to the continual evolution and breakthrough in accelerated computing platforms, such as Graphic Processing Units (GPUs), AI accelerators, neuromorphic computing, and analog in-memory compute.

Occasionally, some players will try to push the envelope from both sides (a good example is Google). However, at GTC Autumn 2022, NVIDIA demonstrated that it is ready to drive AI innovation on both sides through sustained effort in the form of industry-specific solutions.

A key announcement is the launch of NeMo. This is NVIDIA’s first cloud service for developing and deploying large language models, such as the Megatron 530B model. End users can design and train large language models for domain-specific tasks and deploy the models on-premise or in the cloud via NVIDIA’s managed application programming interface services. In October 2022 the service entered private beta mode.

GTC Autumn 2022 also witnessed major announcements in healthcare and robotics. Clara, NVIDIA’s healthcare solution, now features BioNeMo, a service for training and deploying large biomolecular transformer AI models at the supercomputing scale, and IGX, a new-generation AI computing platform for real-time computer vision inference in edge industrial servers. In robotics, developers now have access to Jetson Orin Nano, a newly improved AI inference platform for autonomous robots—following the more powerful Jetson AGX Orin announced in GTC Spring 2022—and Isaac Nova Orin, a configurable computing and sensor reference platform.

These announcements indicate that NVIDIA aims to transition from a “one and done” hardware purchase revenue model to a subscription-based software revenue model. NVIDIA has been actively working to achieve this transition by laying the necessary foundation. In recent years, the company has launched a myriad of industry-specific solutions with a combination of hardware and software, such as Aerial, Clara, DRIVE, Isaac, and Metropolis. While these solutions provide a clear set of optimized hardware and software for domain-specific AI use cases and tasks, the solutions are primarily available as on-premise private deployment.

The introduction of Omniverse in 2020—specifically the Graphic Delivery Network (GDN) at GTC Autumn 2022—is set to change this. Over the years, NVIDIA has partnered with Equinix, the leading data center company, to deploy its DGX GPU platform into Equinix’s 220-plus data centers. In addition, the company is also working with cloud service providers, such as AWS and Google Cloud, to offer its solutions as a form of public cloud. All of these infrastructures grant NVIDIA much-needed convenience and flexibility. They enable NVIDIA to launch new cloud-based services, such as NVIDIA Tokkio for digital avatar development and Maxine for video conferencing, all the while improving developers’ accessibility to NVIDIA’s solution and increasing its touchpoint and user stickiness.

It may be tempting at this stage to call NVIDIA an AI Software-as-a-Service (SaaS) company, but there are several hurdles.

First, revenue from all the industry-specific solutions mentioned above remains low. As of 2021, gaming—standing at 46%—still represents NVIDIA’s largest source of income. NVIDIA continues to benefit from its gaming business, with GeForce NOW, the company’s cloud gaming service, providing the blueprint and the technological foundation for Omniverse’s GDN. In comparison, the revenue from industry-specific solutions is estimated to be around 7% to 10% of overall revenue.

Second, the company may face pushback from its ecosystem partners, specifically the Independent Software Vendors (ISVs) who build their solutions on top of NVIDIA’s hardware and software. Traditionally, these relationships play a critical role in propelling NVIDIA into market leadership in AI because ISVs can confidently develop effective AI software based on NVIDIA GPU and CUDA framework, libraries, and tools. However, the ISVs may be concerned that they have to compete directly against NVIDIA’s own SaaS offerings.

Nonetheless, the enterprise AI market will remain a blue ocean for the foreseeable future. While AI is becoming a priority for senior management, more work needs to be done as most enterprises still require support on AI skill sets; data operations; and developer-friendly frameworks, tools, and libraries. By constantly improving its industry-specific solutions and tackling some of the toughest challenges in specific verticals, NVIDIA intends to push the boundary through its hardware and software innovation.