NVIDIA (NVDA -0.66%) made a slew of announcements this week to kick off its week-long virtual GTC conference, centered on new data center hardware and software capabilities.
It's been a busy year for the pioneer of the GPU (graphics processing unit), which is following up its takeover of networking company Mellanox with the announcement it intends to purchase Arm Holdings for $40 billion from Softbank Group.
If I had to choose only one semiconductor stock to own over the next decade, NVIDIA would easily be it.
Deepening computing processor specialties
First, the hardware announcement. Ahead of a tie-up with Arm, NVIDIA said it will be using the U.K.-based chip designers architecture for a new family of processors for data centers. Back in the spring, CEO Jensen Huang had discussed "the three pillars in modern computing." In Huang's vision, the familiar CPU (or central processing unit, which was invented by Intel) will continue to be used for general purposes. But as Moore's Law loses steam, the GPU invented by NVIDIA will become a computing accelerator, used for specialized data-intensive tasks like AI training and deployment.
But the third pillar of Huang's vision is the DPU (data processing unit), a specialized unit for management of data centers. With the advent of the cloud over the last decade, the data center has become a new basic computing unit. And with access to high-speed internet connectivity on the rise, data centers have become even more important as organizations and individual users tap them for heavy lifting rather than crunching information directly on their local devices.
NVIDIA's new chips, built on Arm's multicore CPU architecture, are called the BlueField-2 DPU family and will start rolling out in server hardware systems in 2021. Paired with the GPU for specialized tasks and the networking gear acquired from Mellanox earlier in 2020, NVIDIA now has a full suite of hardware addressing the modern data center computing unit. Having Arm in the fold would increase NVIDIA's ability to pursue this strategy -- as well as disrupt other markets like smartphones and build on the momentum NVIDIA already possesses in supplying tech for the automotive, manufacturing, medical, and entertainment industries.
Data centers are a future staple of computing
In addition to new hardware, NVIDIA also announced the release of new AI software based on its data center work. High points include natural language processing platform Jarvis, web recommender Merlin, videoconferencing enhancement tool Maxine, and the Omniverse platform for collaborative 3D design and creative work.
What's a chip company like NVIDIA doing in the software realm? Many companies tout the next-gen capabilities of their hardware, but a chip's ability can be seriously handicapped if it isn't easy to integrate within a larger computer system. Software development kits are key to unlocking a piece of hardware's full range of capabilities. The recent releases thus help encourage the adoption of AI systems that use NVIDIA's wares, which in turn boosts sales over time.
With a new addition to its suite of chips and a growing list of ready-to-use software platforms built on it, NVIDIA's traditional slant toward powering high-end computer graphics now lies at the intersection of AI, cloud computing, and other modern computing developments in high-growth mode. If it can bring Arm in-house (or simply doubles down on launching new chips using Arm architecture), NVIDIA says its annual addressable market will be some $250 billion and still growing. For reference, NVIDIA has hauled in just over $13 billion in revenue over the last trailing 12-month period.
This explains why shares carry such a hefty premium (the company carries a market cap of over $330 billion as of this writing). NVIDIA is still in high-growth mode and has plenty of promise in the decade ahead. It will be a wild ride along the way as the chip industry is a cyclical one, but this is quickly becoming the dominant player addressing cutting-edge technology needs.