NVIDIA Competitor Analysis Report
Automotive Industry
NVIDIA Corporation is an American company specializing in visual computing technology…
Founded by Jen-Hsun Huang, Chris A. Malachowsky and Curtis R. Priem in January 1993, industry heavyweight NVIDIA develops and manufactures solutions for visual computing, including graphics processing units (GPUs), system-on-chip units (SoCs), Tegra Processors, computers and software. The GPU products include GeForce for personal computer gaming, Quandro for computer aided design, video editing, special effects, and other creative applications. As a result, their primary market lies in PC gaming as their products are perpetually relevant, but they are broaching new opportunities in automotive and data center end markets owing to their efficacy in solving artificial intelligence (AI) challenges. For example, the DRIVE PX2 facilitates deep learning and autonomous driving, whilst the GRID is a platform for cloud and data centers. NVIDA’s Tegra Processors are primarily designed for its DRIVE, which provide automotive computing for driving and SHIELD, which designs devices to harness mobile cloud power.
NVIDIA & Moore’s Law
NVIDIA is working with automotive partners to develop a robust, self-driving ecosystem. The company is targeting big automotive manufacturers — they recently announced collaborations with Audi and Mercedes, in addition to their previously announced partnership with Tesla. They also work with supplier OEMs such as Bosch and ZF and mapping companies such as TomTom and ZenRen (for 3D mapping)– the goal of which is to develop a robust self-driving ecosystem for the future. As Level 4 driving (hands-free) will not be viable in certain scenarios (due to variables such as location), NVIDIA announced the new AI co-pilot (at CES January 2017) to help the driver when the computer cannot take over driving responsibilities completely. In March, NVIDIA and Microsoft announced a new hyper-scale design for cloud-based AI computing. Gaming catalysts include: competitive gaming, VR/AR, in-game film features and a perpetual upgrade cycle. Even though the company has experienced some decline in growth, they remain the leaders in machine learning in data centers and are therefore positioned for significant future growth in the autonomous vehicle market. Ultimately, they are set for huge leaps in AI and deep learning innovation.
Take their newest product the NVIDIA DRIVE PX2 as an example.
“With advances in GPU computing, fully autonomous vehicles will be in mass production by approximately the year 2020”
DRIVE PX2 is by far the most advanced GPU for deep learning and complex data computations in the field of autonomous driving. The company’s auto-pilot computer combines deep learning, sensor fusion and surround vision to deliver fully autonomous, driver-assist capabilities. The single-processor of DRIVE PX 2 is capable of handling functions like HD mapping and automated driving on just 10 watts of power, making it extremely useful for deploying self-driving AI. The DRIVE PX2 in its premium form is a 4 chip solution consisting of 2 NVIDIA ARM based processors and 2 Pascal architecture-based GPUs, although auto-OEMs can opt for fewer chips in their configurations depending on performance requirements. Furthermore, the DRIVE PX 2 uses deep neural networks to process vast amounts of data from sensors in the car, to understand exactly what’s happening in real-time in the vicinity and therefore safely navigate through traffic. With the advances in GPU computing (resultant of Moore’s law), it is reasonable to assume that fully autonomous vehicles will be in mass production by approximately the year 2020, whilst preceding models will incorporate many advanced self-driving features.
Although prioritizing regulation over innovation could hinder the development of a fully autonomous vehicle, much of the training that deep learning systems will undergo will take place within data centers, where technological advancements are seldom slow. As a result, we can expect to see a continuation in the exponential growth of autonomous driving technology in the next decade. That is not to say that the computing performance within the car itself will diminish due to the use of data centres– it will actually remain high due to the plethora of incoming data from the newly-integrated sensors in each vehicle. This will facilitate real-time decision-making in the vehicle and also provide adequate headroom for future upgrades, as wireless updates of software codes could become increasingly complex and performance intensive. It is also conceivable that manufacturers without centralized, autonomous AI data centers and vehicles lacking wireless capacity may have self-contained machine learning systems that learn behavior and road conditions over time, therefore needing higher levels of performance to execute semi-, or fully-, autonomous features.
While NVIDIA has been strongly supportive of the use of GPUs in deep learning research and applications, Intel’s most recent acquisition of Nervana Systems provides another data point that suggests deep learning is a growth area, and in particular, the importance of hardware semiconductors in the field.
Another major breakthrough (again thanks to Moore’s law) is evident in NVIDIA’s announcement of the JETSON TX2 developer kit, which will be able to provide “computing at the edge” to a wide range of products. The embedded system allows for localized, parallel computing for features such as speech recognition, navigation and other AI tasks without the need for cloud connectivity. This will significantly lower bandwidth usage and security threats. The system is aimed at image processors for automobiles, robotic machinery and drones. The platform offers twice the performance and runs on a lower power mode than previous platforms.
“Moore’s law, cloud computing and GPU innovation will set a new standard for artificial intelligence, which competitors will have to reach to survive”
In conclusion, NVIDIA is transforming the GPU and autonomous driving market, thus positioning itself as a deep learning pioneer. This is evident from their significant investment in R&D in the field of deep learning, as well as the partnerships they are forming with industry leaders in tech and automotive. Even though gaming remains the company’s primary market, we strongly believe the potential of (and demand for) AI will drive the company towards other areas of interest. In the near future, NVIDIA will dominate the market for autonomous driving hardware and provide the foundations for future advancements in AI technology. The combination of Moore’s law, the cloud computing revolution and GPU innovation will set a new standard for artificial intelligence, which competitors will have to incorporate not only to stay relevant, but to survive. NVIDIA will continue to experience growth in the gaming market, data centers and automotive services to continue driving outperformance for the next few years. The company is well positioned to benefit from these growth drivers, and its lack of exposure to traditional PCs, smartphones and other maturing markets leave it positioned to see revenue growth and multiple expansion in the long-term from more than just gaming.
NVIDIA & Metcalfe’s Law
How can a chip producer benefit from network effects? As discussed, NVIDIA is the leader in visual computing and software, in particular graphics processing units (GPUs). Traditionally NVIDIA’s markets lie in PC gaming and professional visualization, but they are moving confidently into data centers (AI) and autonomous vehicles due to pioneering GPU technology. GPUs are rapidly evolving beyond their original design and are poised for rapid extensions linked to AI– specifically embedded vision, autonomous driving and deep learning. In terms of NVIDIA’s market appeal beyond gaming, the turning point was Intel’s staggering $15.5 billion acquisition of Mobileye which has proven their value potential in the autonomous vehicle space.
But does a company like NVIDIA benefit from network effects in the same way as companies such as Uber or Tesla, for example? The answer is– probably not. Uber and Tesla’s monumental successes re: network effects are physically tangible and plain to see. As NVIDIA’s network effects are rooted less tangibly in data centers and AI technology, it is harder to immediately discern where they’re most evident. However, that does not mean they are not present, and effectual. NVIDIA is the world leader in GPU data center acceleration and AI hardware, and targets only 3 niche markets at a valuation of ~$60 billion. These are computer acceleration (HPC), cloud acceleration and enterprise. However, though NVIDIA deploys its GPUs into many applications for these markets, AI and deep learning hold the strongest potential for where network effects will be most evident.
“NVIDIA’s GPU technology will eventually become a standard technology benchmark in the automotive industry”
Every single car manufacturer that wants to excel, remain a market competitor (or even survive) will need to deploy NVIDIA’s GPU technology at some point, which is why the network effects in the fields of AI, deep learning and data centers and guaranteed. It will eventually become a standard technology benchmark in the automotive industry, and can only develop in the same direction. Already, each car running on NVIDIA’s PX2 accumulates vast amounts of data, which is sent to the data centers running on NVIDIA’s GPUs. The volume of data processed, combined with the increasing performance of chips (Moore’s law) results in vast improvements in network strength and value overall. Today, NVIDIA lets car manufacturers handle the data themselves, however this is inevitably set to change in the near future. Since most data centers are run by a small group of tech giants (Google, Amazon and Microsoft), we believe organizations will move towards offloading their data center tasks to cloud providers. The impact of this would proliferate tremendous advances in deep learning, followed by growth in cloud providers.
The performance of deep learning systems like NVIDIA’s PX 2 improves with each new connection as the system experiences and therefore processes, more and more data. By training itself, the company’s deep learning technology can detect, identify and recognize any number of objects. Once this driving data is decentralized, every car on the road will feel the benefits of the system and it’s network– even those cars without the technology installed. Efficiency will improve via a domino-effect for automotive experience as a whole.
NVIDIA & The Power Law
“There are still no competitors on the market that can challenge NVIDIA’s PX line”
NVIDIA’s biggest technological advantage comes from their high-performing GPUs. GPUs enable high performance graphics, and at their most basic level, these graphics represent images made of many small triangles each defined by 3 vectors. To manipulate these images quickly, a computer would need to perform millions of similar operations at once, but as GPUs are specifically designed for such operations, they are far more efficient. They are also well suited for deep learning because the computations required to render an image are similar to those required by deep neural networks. Since NVIDIA is the all-time ‘Category King’ when it comes to gaming, moving towards the field of AI is the natural progression. NVIDIA’s PX products are shipped to over 80 car companies worldwide. They enable center fusion, mapping and detection using just the internet. All this combined has made NVIDIA the leader when it comes to autonomous driving, as every car company can use the open stack PX2 to build their self-driving car on top of it whilst solving multiple issues simply with one chip (breaking, autopilot, etc.).
Even though much of the training that their deep learning systems undergo takes place within data centers (another strong area for NVIDIA), performance computations will still take place within the car. Also, car manufacturers that do not have centralized data centers would require even higher levels of performance. All of this makes NVIDIA’s GPU products for machine learning unique and indispensible due to the performance levels that they achieve across the board. There are still no competitors on the market that can challenge NVIDIA’s hardware products, so even when Intel inevitably comes up with its own solution for deep learning, it would not mean the end of the PX product line.
The only field that NVIDIA has yet to penetrate when it comes to self-driving cars is that of wireless connectivity, which is obviously crucial for allowing a car to connect with its data center and communicate with other vehicles. However, we believe this should not pose an issue for the company once we see those self-driving cars on the road.
