If you have been following Nvidia of late, it seems to offers a dizzying array of products and solutions in industries ranging from self-driving tractors, medical scanning, folding proteins, robots running around, metaverse and of course, everyone’s favourite – Artificial Intelligence (AI).
It still remains the Graphic Processing Unit (GPU) of choice for gamers. But keeping up with the developments in other areas is exhausting.
Is Nvidia taking on too much? Or is it all a part of a carefully thought-out strategy?
Let’s look at Nvidia’s products and solutions to understand the strategy better. We will view this through the lens of:
- Revenue Segments
- Products
- Markets
Revenue Segments View – Graphics & Compute & Networking
Nvidia’s Annual Report boils down all the products and solutions to two segments: Graphics and Compute & Networking.
- Graphics segment
- GPUs: GeForce GPUs for gaming and PCs
- Streaming: the GeForce NOW game streaming service and related infrastructure, and solutions for gaming platforms
- Workstation: Quadro/NVIDIA RTX GPUs for enterprise workstation graphics
- Cloud: Virtual GPU, or vGPU, software for cloud-based visual and virtual computing
- Automotive: Automotive platforms for infotainment systems
- Omniverse: Software for building 3D designs and virtual worlds
- Compute & Networking segment
- Data Center: Platforms and systems for AI, HPC, and accelerated computing
- Networking: Mellanox networking and interconnect solutions
- Automotive: Automotive AI Cockpit, autonomous driving development agreements and autonomous vehicle solutions
- Cryptocurrency: Cryptocurrency mining processors, or CMP
- Robotics: Jetson for robotics and other embedded platforms
- Artificial Intelligence: NVIDIA AI Enterprise and other software.
The Revenues are also reported based on the above two segments.
The revenue bifurcation by ‘Graphics’ and ‘Compute & Networking’ are not very clear on the products and solutions being developed. This view also doesn’t elaborate on the applications of these products and the industries.
Products View – Built like a Full Stack
Another way of viewing Nvidia is like a Computing Stack or Neural Network, with in four layers:
- Hardware: These are typically GPUs and Networking Solutions (mainly from the Mellanox acquisition)
- System Software: This includes:
- CUDA: Nvidia’s invention of Compute Unified Device Architecture (CUDA) in 2006 opened the parallel processing capabilities of GPUs and extended its use to general purpose computing.
- SDKs: Software Development Kits (SDKs) capture the complex computer science and engineering into expertly crafted CUDA libraries
- CUDA-X collection of application acceleration libraries
- Application Programming Interfaces (APIs)
- Domain-specific application frameworks
- Software and Applications: These include applications like:
- Omniverse: software for building 3D designs and virtual worlds
- GeForce NOW: Streaming service for games
- Enterprise Software Products: These are offered on a on a standalone basis as a perpetual license or subscription. They also form the backbone of the Platforms.
- Large Language Models (LLMs): These are used for deployment to customized AI applications. Examples are NeMo LLM and BioNeMo LLM.
- Platforms: Platforms are a combination of the hardware such as Accelerators (GPUs, CPUs, DPUs) and Software such as SDKs and CUDA for a specific purpose. Examples are:
- NVIDIA RTX for Graphics
- NVIDIA HPC for Scientific Computing
- NVIDIA AI for Data Science and AI
- NVIDIA DRIVE for Autonomous Vehicles
- NVIDIA Isaac for Industrial Robots
- NVIDIA Clara Holoscan for AI-powered medical systems
- NVIDIA Omniverse
Markets View: The 4 Key Markets
Nvidia operates in 4 key markets and this view makes the most intuitive sense: Gaming, Data Center, Professional Visualization & Automotive (or Robotics).
1. Gaming – the Bread and Butter business
The Gaming market for Nvidia is primarily products/hardware such as GPUs and Streaming Services.
- Gaming desktop and laptop PCs are powered by GeForce RTX and GeForce GTX GPUs.
- In Streaming, there is GeForce NOW cloud gaming for playing PC games over the Internet and SHIELD for high quality TV streaming.
2. Data Center – And this is where ChatGPT and other AI Models come in
Nvidia’s Data Center products are a great example of the full stack approach.
- The hardware layer consists of GPU & DPU chips. These chips are scalable so you can stack them together to create bigger computing units.
For example, Eight A100 chips provide the basis of a DGX system.
Nvidia also uses CPUs from Intel and AMD but plans to develop it’s own CPUs for Data Center applications - Interconnects And Systems: This includes Networking Systems from Mellanox and Software/SDKs etc that we discussed in ‘System Software’ above.
While currently Nvidia mentions Data Center as a market, we can add Artificial Intelligence (AI) systems to this category. Most Large Language Models (LLMs) are built on the Cloud.
In Nov 2022, Nvidia entered into an agreement with Microsoft Corp (MSFT.O) to build a “massive” computer to handle intense artificial intelligence computing work in the cloud.” This is path breaking as it’s the first time Nvidia’s full Artificial Intelligence (AI) stack (composed of Nvidia hardware, software and interconnects outlined above) is being used.
With the proliferation of ChatGPT (text AI), DALL-E (image AI), we are now seeing an arms race for acquiring AI capabilities. Nvidia’s AI business should stand to gain – to the extent that we might see this as a new market being reported separately.
‘Including GPUs and networking, NVIDIA powers over 70%, and 8 of the top 10, supercomputers on the global TOP500 list.’
Source – 2022 Annual Report, Nvidia
3. Professional Visualization – Where Nvidia takes on Facebook’s Metaverse
Nvidia describes its offerings in the Professional Visualization market as follows:
- NVIDIA RTX platform: makes it possible to render film-quality, photorealistic objects and environments with physically accurate shadows, reflections and refractions using ray tracing in real-time.
- NVIDIA Omniverse: is a virtual world simulation and collaboration platform for 3D workflows that is available as a software subscription for enterprise use and free for individual use.
Facebook/Meta’s approach is to replicate Facebook in the Metaverse. Nvidia’s Omniverse goals are different. It is currently focusing on Industrial applications such as digital twins of factories, simulation environments for robots and automobiles to self-learn & 3D design and rendering.
In some aspects, its ambitions are truly astounding – it aims to create a digital twin of the Earth itself – no less! It’s called the Earth Observation Digital Twin (EODT) and aims to simulate weather and other natural phenomena.
4. Automotive or Robotics – In a race with Tesla and Elon Musk
This comprises cockpit & AI solutions including autonomous driving. Nvidia DRIVE is the platform that offers this end-to-end solution for the AV market.
In Jan 2023, Mercedes surprised everyone by becoming the first auto company to offer Level 3 Autonomous Driving in Nevada US – beating the mighty Tesla in the process. It had done the same in Germany a few months earlier.
What the breathless reporting missed was that the ‘brains’ behind the Autonomous Driving was Nvidia’s DRIVE Pilot. Some experts consider Nvidia’s solution to be superior to Tesla’s as it also gets inputs from LIDAR. There are costs and benefits to Tesla’s approach vs Nvidia’s and the jury is still out.
Nvidia-powered robots are making giant strides
Robotics is another exciting area were Nvidia has been active in the past year. Like Nvidia Drive for Automotive segment, the heart of Nvidia’s Robotics and Edge systems lies in NVIDIA Orin.
The new Orin robotics processor chip is the central computer for a new generation of logistics robots, delivery robots, and self-driving cars, trucks, and taxis.
Nvidia Annual Report, FY22
The use cases are: ‘The NVIDIA Orin robotics computer powers Isaac, Holoscan, and DRIVE. Orin is in production and already powering 25 electric vehicle makers building autonomous vehicles.’
Nvidia’s Strategy Emerges
Nvidia’s core strength is the ability to develop hardware (GPUs) with the fastest computing power, quarter after quarter, for decades on end. Remember, semi-conductors is one of the most fiercely competitive industries in the world. Andy Grove of Intel famously said that only the paranoid survive.
In some ways, Moore’s Law is a standard in the industry – and that requires you to double the computing power every 2 years. So just to keep up, chips have to be 41% more efficient every year.
Platform Strategy to the Rescue
Nvidia operates in this environment and to consolidate its lead, it emphasizes on building a ‘platform strategy’ or a full-stack approach.
This is clearly articulated in its 2022 Annual Report:
NVIDIA has a platform strategy, bringing together hardware and systems, software, algorithms and libraries, and services to create unique value for the markets we serve.
While the computing requirements of these end markets are diverse, we address them with a unified underlying architecture leveraging our GPUs and software stacks. The programmable nature of our architecture allows us to support several multi-billion-dollar end markets with the same underlying technology by using a variety of software stacks developed either internally or by third party developers and partners.
The large and growing number of developers across our platforms strengthens our ecosystem and increases the value of our platform to our customers.
Nvidia Annual Report, 2022
A Natural Progression – May the Fastest Chip Win
Nvidia’s forays into newer areas are a natural progression to its Platform Strategy. Its development of CUDA and acquisition of Mellanox for networking hardware were vital for maintaining its supremacy in computing capability.
Given the intense demand for Accelerators (GPUs, CPUs, DPUs) for High-Performance Computing (HPC), expanding to Data Centers was the logical outcome. The explosion of Artificial Intelligence (AI) led to unprecedented demand for 2 things Nvidia had in plenty; data storage hardware and insanely fast computing.
The development of the Omniverse, a metaverse environment with primarily industrial applications, builds on Nvidia’s raison d’etre – building ‘photorealistic objects and environments’. The original application was for gaming. Here Nvidia pioneered Ray Tracing & ‘machine-learning based super-sampling’ (DLSS 3) – innovations that make games look almost real. The Omniverse is a natural culmination of this journey.
The arrival of 5G and Edge devices opened up possibilities of intelligent machines and Robotics. Nvidia now uses its Omniverse to first train these robots in a virtual world (Isaac Gym). The brains of these pre-trained robots are Orin Robotics Processor Chip. Orin provides the ‘intelligence’ for logistics and delivery robots. It can also manufacture cars – ‘already powering 25 electric vehicle makers building autonomous vehicles’.
The autonomous vehicles made by Orin will mostly have a Nvidia DRIVE as the cockpit management system and Autonomous Driving Solution. While revenues from Automotive are still around $0.5 billion in a year, the pipeline is exciting with $11 billion ‘design win pipeline over the next six years’.
Conclusion – The Greed for Compute Power
Today’s world of Data Centers, Artificial Intelligence (AI), Metaverse-like environments, 5G enabled advances in Robotics, intelligent cars and ‘Intelligent Everything’ requires raw computing power. It cannot get its hands on enough computing ability that Nvidia provides.
Sure, there are all kinds of competitors – the traditional chip companies like AMD, the giants who wish to develop expertise in-house such as Tesla and Google and hundreds of AI start ups like Cerebras. And breakthroughs in science such as Quantum Computing looming on the horizon.
But remember, Nvidia invented the modern GPU and it has stayed the fastest GPU in town for much of its existence. With a 65% gross margin, there is no shortage of competitors at the gates. And yet the only ones to beat Nvidia’s chips are Nvidia’s new chips.
And so we expect Nvidia to become an integral part of an exciting new world.