Nvidia's Potential Is Massive as 2 New Data Center Types Emerge
By Nicholas Rossolillo
Jun 1, 2022
KEY POINTS
- CEO Jensen Huang talked about two new data center types Nvidia hardware is helping to create.
- With AI proliferating across the economy, Nvidia has massive potential ahead of it over the next decade.
- A new breed of data centers is emerging that needs high-performance chips running around the clock.
Nvidia ( NVDA -1.88%) reported record sales from data centers during Q1 (the three months ended May 1, 2022). In fact, as AI becomes more useful and affordable, it's no longer the realm of high tech. Industries across the entire economy are putting AI to work, and Nvidia is emerging as the top platform for the movement. As a result, data centers are likely to remain the company's top revenue source going forward.
This transformation of Nvidia's business means there could be many more years of rapid growth ahead. CEO Jensen Huang explained on the earnings call that two new data center types are emerging, and construction of these AI computing units is only just beginning.
IMAGE SOURCE: GETTY IMAGES.
The rapid emergence of a $10 billion-plus per year business
As semiconductor technology improves computing power and energy efficiency, a new hardware upgrade cycle kicks off every few years or so. All physical infrastructure, chips included, have a finite life expectancy. The nice thing about technology infrastructure (like data centers that power cloud computing services) is that when a company starts replacing old equipment, it's getting an upgrade in performance too.
An upgrade cycle like this is currently underway, set off by the pandemic and an increased focus on digital tools after the world went on temporary lockdown in 2020. To maximize its take from this upgrade cycle, Nvidia has been devising new chips aimed at the data center and AI spaces. The company's transformation from high-end video game chip designer to AI platform has been stunning over the last few years.
With Huang and company anticipating sequential growth for the data center segment every quarter for the rest of this year, it appears data centers will be Nvidia's largest end market going forward (data centers briefly overtook the gaming segment for a single quarter a couple years ago before consumers went on a PC upgrade cycle of their own at the start of the pandemic).
FY 2017 FY 2022. Q1 FY 2023|
| Video Game Segment | $4.06 billion | $12.5 billion | $3.75 billion | | Data Center Segment | $830 million | $10.6 billion | $3.75 billion |
DATA SOURCE: NVIDIA. NOTES: FISCAL YEAR 2017 ENDED JANUARY 2017, FISCAL YEAR 2022 ENDED JANUARY 2022, FIRST QUARTER FY 2023 ENDED MAY 2022.
Video gaming sales have been impressive in their own right the last few years, but Nvidia has become an incredibly powerful player on the data center front. It hauled in as much data center revenue in just a few weeks last quarter as it did the entire year of fiscal 2017. For the sake of reference, the current hardware upgrade cycle could keep revenue climbing. Intel pulled in $6 billion in sales from its "Data Center and AI Group" segment during the first quarter. Basically, Nvidia has room to gobble up lots more market share from Intel in this department.
Not just a hardware upgrade cycle
But this isn't just a story about data center equipment upgrades. Nvidia's AI technology work has helped spur on the emergence of two new types of data centers (beyond the traditional use case as a cloud services computing unit). Huang explained on the earnings call:
Over time, you're going to see even more types of [chip] configurations. And the reason for that has to do with a couple of very important new types of data centers that are emerging. And you're starting to see that now with fairly large installations, infrastructures with Nvidia [high-performance computing] and Nvidia AI. These are really AI factories where you're processing the data, refining the data, and turning that data into intelligence. These AI factories are essentially running one major workload and they're running at 24/7. Deep recommender systems is a good example of that...
And then on the other end, you're seeing data centers at the edge that are going to be robotics or autonomous data centers that are running 24/7. They are going to be running in factories and retail stores and warehouses, logistics warehouses, all over the world.
The first of the two new data center types Huang mentioned is "AI factories," built to ingest massive amounts of data to train AI models (software systems that will later be used to make decisions, recommendations, or automate a process). We already interact with the products of these "factories," like when a website recommends items we might be interested in purchasing. Facebook parent Meta Platforms announced one of the world's most powerful AI factoriesearlier this year with the help of Nvidia hardware.
The second type of new data center is essentially an operator of robotic equipment. This could be an automaker factory building cars, or a retailer's distribution center processing and shipping customer orders. A few years ago, the U.S. Postal Service started working with Nvidia to produce an AI platform to help make the massive mail system more efficient. This is another example of a robotics data center operating at the "network edge."
To sum it up, Nvidia's AI hardware is helping it gain market share against its peers in the existing data center space. But Nvidia is also helping expand the scope of the industry, putting its business in a unique position to create its own destiny. It's hard to quantify how big an opportunity Nvidia still has ahead of it, but judging by the momentum the company has been riding over the last few years, AI data centers are going to be a very big deal by the end of this decade.
fool.com |
