last posts

Will 2023 be the year Nvidia's software business really takes off?


A dreaded cyclical downturn for the chip industry has arrived. In concrete terms, consumers are holding back the purchase of new PCs and smartphones, which has hit Nvidiait is (NVDA 0.07%) empire built on the particularly tough GPU (graphics processing unit). During the last quarter, sales of video games and other graphics applications were halved compared to the previous year.

Enterprise computing, however, is a whole different story. Right now, the sale of actual GPUs is still fueling strong growth at Nvidia, and a budding software company is also starting to emerge. Will 2023 be the year Nvidia moves from the semiconductor business to software?

Will Nvidia software revenues take off in 2023?

To be clear, Nvidia doesn’t separate the revenue it makes from software services from the revenue from the sale of actual semiconductor hardware. We can assume, however, that the bulk of sales still come from the sale of GPUs and GPU-powered computer systems for data centers. After all, Nvidia has a long history of bundling freeware with a chip purchase to get gaming enthusiasts and developers up and running with their new GPUs as quickly as possible.

However, Nvidia is already making more money from software sales than many investors realize. For example, many Nvidia chips are used for incredibly complicated computing tasks that require advanced software programs to run. Many companies using Nvidia hardware aren’t software technologists themselves, so using some of Nvidia’s software engineering is already happening. Some of that software access may be a one-time purchase, rather than the recurring software-as-a-service (SaaS) solution that investors love.

As evidence, during the third quarter FY2023 earnings call, CFO Collette Kress explained that Nvidia would begin segmenting its large “data center” unit. It is no coincidence that “data centers” are by far the largest unit of the company. It had more than $3.8 billion in sales (more than half of total revenue of $5.8 billion) in the quarter. Kress said:

As the number and scale of public cloud computing and Internet service companies deploying NVIDIA AI grows, our traditional hyperscale definition will need to be expanded to reflect different end-market use cases. We will align our customer feedback on data centers accordingly. Other industry verticals, such as automotive and energy, also contributed to the growth with key workloads related to autonomous driving, high performance computing, simulations and analytics.

In other words, “data centers” have become far too generic a term for Nvidia’s largest revenue segment. This big $3.8 billion quarterly item included actual GPU sales to data center operators, GPU “rentals” through a cloud service (called Infrastructure-as-a-Service, or IaaS), services engineering and software used to operate. these GPUs. Starting with the next earnings report, expect plenty more detail from Nvidia about what its customers are actually doing with its silicon designs.

Building from small beginnings, but how much?

Sales of Nvidia’s data center unit are currently a little opaque, but what about the other three main end markets of “gaming”, “professional viewing” and “automotive”? There are also no details on software sales here, but Nvidia could already be making a lot of money from recurring software revenue.

Let’s start with Nvidia’s first subscription business, GeForce Now (think of it as netflix (NASDAQ:NFLX), but for video games). Last summer, Kress said registered GeForce Now users topped 20 million worldwide. How much revenue does it generate? Here’s a little guess: GeForce Now has a free tier (I guess half of the 20 million users subscribe to it), a “Priority” pass at $9.99 per month (say nine million subscribers on average, or $90 million in revenue) and an Advanced Pass at $19.99 per month (one million subscribers or $20 million in revenue).

If this is even a remotely accurate guess, it means Nvidia could generate over $300 million in revenue on a quarterly basis. In the third quarter of fiscal 2023, total video game revenue was $1.57 billion, so video game streaming could still be a small – but certainly not insignificant – source of sales rounding out the sale of video game GPUs.

Something similar likely happens with the “professional visualization” segment, which caters to video game and movie content creators, engineers, architects, and more. Revenue was just $200 million last quarter, a sharp drop due to lower PC sales. However, Nvidia’s Omniverse Cloud subscription (used to create 3D virtual worlds, for games and movies, or to simulate real-world locations) is bundled into this unit.

Omniverse is free for individuals, but a basic enterprise package starts at $9,000 per year. Hundreds of companies now use Omniverse, so it wouldn’t take many teams subscribing to the service to make it a big part of the “business visualization” segment.

The “automotive” segment is even more delicate. Sales were $251 million last quarter, and we can assume that most of that revenue comes from selling Nvidia’s computer system used in modern cars (Nvidia DRIVE Orin and Thor chips). However, Nvidia offers a whole suite of software for companies developing autonomous vehicle technology, from actual computer chips to software used to accelerate development. However, many of these sales could be housed in the “data center” segment. Look for more details on this starting in the fourth quarter of fiscal 2023.

Either way, Nvidia has plenty of irons in the fire when it comes to software, and 2023 could be a big year. But even now, it looks like the company has been quietly expanding its recurring software sales to complement its core GPU designs. As more and more companies adopt Nvidia chips to develop their own artificial intelligence systems and 3D applications, there is huge potential for Nvidia software to continue to grow. This is a truly unique company operating within the massive semiconductor industry.



Font Size
lines height