All Eyes On The Intelligent Edge At NVIDIA GTC 2022

By Patrick Moorhead - October 13, 2022
NVIDIA CEO Jenson Huang at GTC Fall 2022NVIDIA

Last week, NVIDIA held its annual GTC conference for AI developers, engineers, researchers, IT professionals, industry insiders, and everyone in between. As the company sits firmly in my purview as a tech analyst, I always attend in-person or virtually (see my coverage of 2021’s events from here and here). Whereas last year I focused heavily on NVIDIA’s automotive news, this year I wanted to devote coverage to the company’s big announcements around the intelligent edge. From new Jetson Orin Nano system-on-modules for entry-level edge AI and robotics to new reference platforms for autonomous mobile robotics (AMR) to inroads into medical edge and robotic applications, NVIDIA is gunning for edge AI leadership. Let’s take a closer gander at what the company announced and what it could mean for the tech industry, businesses and consumers alike. 

Jetson Orin Nano

The two new Jetson Orin Nano system-on-modules claim up to an 80x performance over the previous generation. This is big, big, big for developers. The first module, the 8GB version, promises AI performance upwards of 40 TOPS (trillion operations per second) and is power configurable from 7W (watts) to 15W. An even smaller 4GB version offers up to 20 TOPS and is configurable as low as 5W to 10W. I like NVIDIA’s approach to pumping a bit more power into the configuration while getting a lot more performance. Targeting the entry-level tier of edge AI and robotics, the new Orin Nano modules represent the smallest Jetson form factor yet. Specifically, NVIDIA cited Orin Nano’s potential for retail analytics and industrial quality control applications, given its ability to perform complex AI models at a lower price point. 

Notably, the Orin Nano modules are both form-factor and pin-compatible with the earlier Orin NX modules, allowing customers to begin developing for Orin Nano immediately using the AGX Orin devkit. Additionally, one system can support multiple Jetson modules, making it easier for customers to scale their applications. In this uncertain world of IoT, that’s important to keep development costs lower. 

Altogether, NVIDIA’s Jetson family now spans six Orin-based production models, targeting a full range of size and performance requirements—from the new Orin Nano modules on the low end to AGX Orin, capable of up to 275 TOPS on the high end. All these Orin modules sport an NVIDIA Ampere architecture GPU, Arm-based CPUs and next-generation accelerators for deep learning and vision. The modules also feature high-speed interfaces, fast memory bandwidth and multimodal sensor support. According to NVIDIA, the performance and versatility of Jetson Orin empowers developers to create solutions at the edge previously thought to be possible. Families of products are important in this space as customers, again, can leverage development costs across many different end product models. 

NVIDIA also highlighted the broad ecosystem and software support Jetson Orin has already garnered a mere six months since its initial availability. Jetson AGX Orin’s 1,000+ customers and 150+ partners include luminaries such as Canon, John Deere, Microsoft Azure and many more. According to Deepu Talla, VP of embedded and edge computing at NVIDIA, Orin Nano only stands to grow this adoption. All of this adds to the NVIDIA Jetson ecosystem at large, which claims over one million developers and 6,000 customers (a third of which are startups).

Introducing NVIDIA IGX for medical edge AI

Let’s move on to NVIDIA’s new IGX hardware and software platform for medical edge AI use cases. Powered by NVIDIA IGX Orin, IGX promises to provide secure, low-latency AI inference capabilities for clinical settings, which require near-instant insights from a wide range of medical devices and sensors. 

Notably, IGX supports NVIDIA Clara Holoscan, a domain-specific platform that enables developers of medical devices to bridge edge, on-prem and cloud services. According to NVIDIA, this support allows developers to quickly develop new, software-defined devices that can bring cutting-edge AI applications to the operating room. This value prop appears to have struck a chord in the industry. Three leading medical device startups, Activ Surgical, Moon Surgical and Proximie, have already chosen Clara Holoscan on NVIDIA IGX as the foundation for their development of surgical robotics systems.

Boston-based Activ Surgical, for its part, is leveraging NVIDIA Clara Holoscan on IGX to speed up the development of an AI-augmented reality (AR) solution for surgical guidance in real time. Meanwhile, Paris-based Moon Surgical is leveraging the platform to develop Maestro, an accessible adaptive surgical-assistant robotics system designed to work with the existing equipment and workflows in the operating room. Moon Surgical says NVIDIA’s platform has saved it time and resources, allowing it to accelerate its development timeline. Lastly, Proximie, based in London, is leveraging NVIDIA Clara Holoscan on IGX to build a telepresence platform for real-time, remote surgical collaboration. 

Edge AI can accelerate time to diagnosis and save lives. It’s nice to see NVIDIA participating in this market but I wish the industry would speed up its adoption as lives depend on it. I have seen some very interesting startups adopting a “whole body scan” for preventative health. This is the future of diagnosis and preventive care.

NVIDIA Isaac Nova Orin for AMRs

Lastly, NVIDIA revealed the details of three reference platform configurations for its Nova Orin autonomous mobile robot (AMR) platform. AMRs are essentially autonomous vehicles for unstructured environments instead of roadways. Capable of navigating around obstacles and requiring no pre-programmed route or track, fleets of AMRs are already in use in a variety of logistical settings. They’re well suited for tasks such as moving items around warehouses, distribution centers and factories, or for hospitality, cleaning or roaming security applications. 

Two of the new reference platform configurations leverage a single Jetson AGX Orin. Of these, one features safety-certified sensors and a safety programmable logic controller, while one goes without. The third reference platform configuration employs two Orin modules and leverages vision AI to enable functional safety. All designs include support for stereo cameras, lidars, ultrasonic sensors and inertial measurement units. NVIDIA says it chose these sensors taking into account performance, price and reliability for industrial applications.

While AMR manufacturers have historically sourced and integrated computing hardware, software and sensors in-house, NVIDIA says NOVA Orin will eliminate this legwork by providing customers with tested, industrial-grade sensor configurations, software and GPU-computing capabilities. This solution, according to NVIDIA, will allow developers to instead focus on developing unique software stacks of robotic applications. 

If you’re wondering what’s behind NVIDIA’s push into AMRs, look no further than the projected market growth for intralogistics enabled by AMRs—ABI Research expects the market to grow to a whopping $46 billion by 2030 (nearly a six-fold increase from 2021’s $8 billion). With all the work NVIDIA has already invested in autonomous vehicles, this is a very logical (and likely profitable) foray. 

Wrapping up

NVIDIA is a 400-pound gorilla in the tech industry, especially regarding AI training and inference. The thing about a 400-pound gorilla is it pretty much does whatever it wants, whenever it wants to. When NVIDIA decides to throw its weight into the ring, as it has done with the Jetson lineup in the edge AI market, it holds an inherent advantage in terms of size, technology, experience and pre-existing partner and customer relationships. 

Beyond that, I believe the company is wise to offer a full range of edge AI solutions to businesses. AI can be intimidating for those who haven’t worked with it before. The Orin Nano system-on-modules should provide a friendly, entry-level jumping-off point for customers while delivering unprecedented levels of AI performance. Further, the scalable nature of the Jetson Orin modules makes it that much easier for businesses to continue building out their applications once they’ve gotten their toes wet. 

I look forward to seeing more customer and partner stories on how they are leveraging NVIDIA’s IGX platform and Clara Holoscan in the medical device industry. Hospitals and emergency rooms are still struggling to retain staffing, maintain a high care standard and bounce back from the Covid pandemic. The incorporation of real-time, secure, AI-capable devices could do much to improve outcomes and reduce patient wait times in a variety of clinical settings. The Nova Orin AMR reference architectures are also very exciting when you think about how they could accelerate time-to-market for new, potentially transformative AMRs across industries.

All in all, it was an exciting, informative GTC. I look forward to seeing what next year’s event brings and what edge AI inroads NVIDIA makes between now and then.

Patrick Moorhead
+ posts

Patrick founded the firm based on his real-world world technology experiences with the understanding of what he wasn’t getting from analysts and consultants. Ten years later, Patrick is ranked #1 among technology industry analysts in terms of “power” (ARInsights)  in “press citations” (Apollo Research). Moorhead is a contributor at Forbes and frequently appears on CNBC. He is a broad-based analyst covering a wide variety of topics including the cloud, enterprise SaaS, collaboration, client computing, and semiconductors. He has 30 years of experience including 15 years of executive experience at high tech companies (NCR, AT&T, Compaq, now HP, and AMD) leading strategy, product management, product marketing, and corporate marketing, including three industry board appointments.