NVIDIA Is Not A Chip Company. It’s A Platform Company.

by | Nov 9, 2021 | In the News

NVIDIA extends its lead in accelerated computing, announcing platforms for the metaverse, healthcare, enterprise AI, autonomous vehicles, and data center security at the GTC Conference.

NVIDIA, probably on its way to becoming a trillion dollar company, has been building platforms tailored to specific markets, while its competition is still just trying to get competitive silicon to market. Yes, startups and giants alike are bringing out some pretty competitive chips, such as the recently previewed AMD Instinct MI200 and Intel Ponte Vecchio. However, NVIDIA is far ahead of all competitors in terms of its solution portfolio and ecosystem. Consequently, none of these impending competitive threats will materially take business away from NVIDIA AI until competitors’ software ecosystems emerge to turns cool silicon into real solutions. And that will be a difficult and expensive undertaking. (HPC, on the other hand, is a different story, where HPC applications can be ported by ISV’s seeking to sell to supercomputing centers.)

What did NVIDIA announce?

CEO Jensen Huang kicked off the event, as usual, with a slew of announcements, but this time the solutions were the star of the show, not a shiny new chip. (But, there were new chips as well!) We cannot possibly cover all the announcements made at GTC in a blog, but here’s a few worth noting.

Omniverse Enterprise

One area Jensen has been highlighting for a while is NVIDIA Omniverse, the multi-media 3D collaboration platform that enables companies to collaborate in a metaverse using physically accurate “digital twins” of the product being developed or problem being simulated. It could be a car, a factory, the planet Earth, or a new Hollywood blockbuster. Jensen announced that Omniverse Enterprise, previously in beta, is now available from global system vendors, starting at just $9K per year.

So, consider this: Facebook just became Meta, in name only for now, while NVIDIA has been creating and testing a metaverse platform (hardware, software, AI models and services) that increases productivity, collaboration, and accuracy of content creators, designers, and engineers. Real jobs, not future socializing. Why do robbers rob banks? Because thats where the money is. Same situation here. Not to be left out, Microsoft recently announced their own Metaverse platform, Mesh for Teams, to enable an immersive collaboration environment, but that is not really comparable to Omniverse. In the emerging market for the metaverse, NVIDIA is unquestionably the leader.

Jensen’s goal is to reach 40 million 3D developers with Omniverse. NVIDIA has worked with partners to include content and applications from third parties, including ESRI, PTC, Reallusion and Pixar for the Universal Scene Description. As for hardware, 10’s of millions of NVIDIA RTX GPU users will be able to use Omniverse. (Hololens and Oculus headsets not required.)

NVIDIA also announced a new Avatar Creator, part of Omniverse Enterprise, and provided this shot of Jensen and his toys in an Omniverse. Nice Jacket.

We believe Omniverse is the poster child for the dozens of platforms coming from NVIDIA, and is already integrating with other NVIDIA platforms such as the updated Maxine virtual avatar platform, which uses the NVIDIA Unified Compute Framework that unites NVIDIA platforms.

Autonomous Vehicles

Pervasive, end to end support for vehicle manufacturers is another, yes, platform that provides data collection, mapping, training, simulation, driving, and an AV world model. This has now been extended with a synthetic replicator to alleviate the data labelling burden required for training AVs. A new Drive Concierge supports Automatic parking, Summoning, Digital assistant, and Confidence view in the vehicle. And of course, there is new hardware covered below.

New Hardware

Ok, there was some hardware announced as well. Jetson is getting an upgrade to the Orin SOC that is in the Drive platform for AVs, and boasts 200 Int8 TOPS with an Ampere GPU, the Deep Learning Accelerator, and 12 Arm cores. NVIDIA Jetson has had a great reception in the robotics market, and is now used by over 850K developers in over 6000 companies.

NVIDIA announced the new Drive Hyperion 8, combining 2 Orin SoCs with qualified sensors from four leaders, with support for 12 cameras, 9 radars, 1 lidar, and a dozen ultrasonic sensors. While Teslas gets a lot of limelight, NVIDIA boasts a full end-to-end platform (there’s that word again!) for autonomy and over 16 vehicle design wins, now including Lotus.

For the healthcare market, NVIDIA launched a new Clara Holoscan platform with an Arm / RTX A6000 GPU and a ConectX DPU. NVIDIA made several other announcements for Healthcare as well.

Speaking of DPU’s, NVIDIA announced another platform (Morpheus) that uses the smart network interfaces and software to create Zero-Trust Data center protection with BlueField. In a modern data center, each server can be considered a threat. So the firewall protection concept must be taken from the perimeter to each and every server.


NVIDIA, whose stock was up some 24% last week before losing a little ground on Friday, and up over 300% in the last 12 months, is already transitioning to a platform company just when competitors are vying for attention with chips. We believe these platforms will further strengthen NVIDIA in the heterogeneous compute landscape and broaden their reach considerably.

I saw a company last week tout their advantages for edge AI. They have decent performance/watt, to be sure. But comparing to Jetson AGX is like comparing oranges to orangutans: AGX is a platform, with GPUs, DLAs, CPUs, and a rich software stack. The competition has a chip. Platforms are easier to deploy, reduce development time and expense, and enable a carrier-grade solution. (Insert Mic Drop here.)