The Business Of
Semiconductors | Nvidia | How Do They Make Money?
Morning All!
So yesterday, we started our new series on The Business Of Semiconductors. And we kept things fairly high-level, looking at the industry as a whole. Today, we’re going deeper and tucking into Nvidia specifically!
We mentioned yesterday that Nvidia is a designer of chips. But what kind of chips do they design? And what industries use their chips? Well, the chart below gives us the answer to that second question. Nvidia’s primary markets are data centers and gaming!
Now, what on Earth does Nvidia mean when they say they make $24 billion in revenues from gaming and data centers? Well, that’s what we’re going to dissect today!
And just before we do, let’s look at what type of chips Nvidia make. Remember, we had logic (CPU and GPU), memory (DRAM and NAND) and analog. Well, the chart below shows us that 87% of Nvidia’s revenues come from GPU chips! And 13% from their tegra processor, which is a CPU chip. Which means that Nvidia produces solely logic chips.
Alrighty, that’s enough of an intro. Let’s dive in and see how exactly Nvidia makes money!
Feeding Gamers With Graphics
Okay, so let’s start with gaming. Because this was the main bulk of revenues till recently. And when Nvidia talks about gaming revenues, what they’re actually talking about is PC gaming. Now, I’m not a huge gamer, and I’m aware many readers may not be either. So I’ll try and keep this super simple!
So, there are two main products that Nvidia’s involved with in gaming: (i) graphic cards, and (ii) gaming laptops. Let’s start with graphic cards. And first question… what on earth are graphic cards?!
Well, to answer that, let’s first briefly touch on CPU and GPU chips. Most PCs and laptops are built such that the CPU chip comes with a fully integrated GPU chip. And lots of computing applications can run fine on these integrated systems.
However, for very resource-intensive applications (like PC gaming), we need separate GPUs, or discrete GPUs. To lighten the load on the CPU chip. And that’s where graphics cards come in! These graphics cards are boards (called add-in boards) that incorporate the GPU chip. And when it comes to the graphics cards in gaming PCs - Nvidia is a huge player! The graphic below gives a nice (but small and slightly hard to read!) illustration of how it works…
Now, I said Nvidia is a huge player. And that’s maybe a slight understatement! Nvidia actually has over 80% market share in the discrete GPU market. The only real competitor in this market at present is AMD. Making this market a virtual duopoly!
And despite their efforts, AMD are actually falling further behind. Nvidia’s GPU chips are well-known for providing gamers with the best graphics. And so when graphic card manufacturers are deciding which chips to go for. They know having Nvidia’s chips will help them sell their graphic cards to gamers!
Okay, so the 80% market share - what does this mean in terms of sales? Well, last year, Nvidia sold 30 million discrete GPUs to the add-in board (AIB) manufacturers. But how much are they selling these GPUs for? Well, a glance on the company’s website shows us that these graphics cards range in price from $589. All the way up to $1,999! (I had no idea these cards were so expensive!) Now, these listed figures are what the AIB manufacturers receive. What Nvidia sells the GPU to the AIB manufacturer for is more uncertain. But let’s estimate an average of $200. 30 million GPUs sold x $200 per GPU = $6 billion revenue. This feels about right considering total gaming revenues are ~$9bn!
Right, so that’s the graphics cards. What about the gaming laptops? Well, it’s a similar story so I won’t spend too long here. Nvidia designs the GPU chips that power these gaming laptops. And the screenshot below shows us the example of the ASUS ROG gaming laptop. The first orange circle shows us that the GPU chip powering this laptop is Nvidia’s GeForce RTX GPU. And the CPU chip (second circle) is the Intel Core chip.
So, designing and selling the GPU chips in graphic cards and gaming laptops. That’s how Nvidia makes its money in the gaming segment! I found it pretty crazy how Nvidia could make so much money from the gaming market. But when you realise that there are ~3 billion gamers in the world. And ~2.9 billion active users on Facebook, it starts to make more sense!
And the huge growth in PC gaming, driven by the rise of esports, live streaming and high-quality graphics. Has led to Nvidia’s gaming revenues performing very strongly over the last decade. From $1.5bn revenue in 2014, Nvidia’s gaming segment brought in $12.5bn revenues in 2022. Before macro issues weighed on 2023’s performance.
Supercomputers = Super-revenues!
Okay, so we’ve covered gaming revenues. And as I mentioned, for all of Nvidia’s history, gaming was where the company made most of their money. However, this has been gradually changing. And in 2023, the data center segment actually became the primary revenue generator!
But what exactly is this data center segment? Well, if the price of the ASML machines didn’t blow your mind yesterday, I’m pretty certain this will now. Let me set the scene. In around 2007, Nvidia realised that the GPUs that they were designing for gamers could actually be used for far more complex applications. Applications such as AI and machine learning. And over the last year, we’ve seen companies from all industries look to develop their AI/ML capabilities. A very recent example is Open AI and Chat GPT…
Okay great, but where does Nvidia come into this… ? Well, you know yesterday we mentioned that the semiconductor chips in our phones need more and more transistors on them. To handle the increasing complexity of phones. Well, compare the complexity of phones to the complexity of large language models like ChatGPT! Models like ChatGPT are so much more complex, and require so much more data to be processed. That they need hundreds and thousands of GPU chips!
In fact, earlier this year Microsoft announced that they had built a new supercomputer specially for ChatGPT…
And in order to build this supercomputer, Microsoft used 10,000 GPUs from Nvidia! But the crazy thing is that each GPU from Nvidia cost Microsoft $10,000. This means that to build the supercomputer, Microsoft paid $100 million to Nvidia just for the cost of GPU chips!
And it doesn’t stop there! In 2021, Tesla unveiled their supercomputer powered by Nvidia. For a company focused on something as complex as autonomous driving, it’s no surprise that they need lots of GPU chips. The Tesla supercomputer has a total of 5,760 GPUs from Nvidia. Meaning Nvidia would have made around $57 million from this deal (5,760 GPUs x $10,000 per GPU = $57m)! Incredible figures.
But maybe the most extraordinary stat in all this is that Nvidia’s share of the enterprise GPU market is a mind-boggling 91%! So not only does the company operate in a duopoly in the PC gaming GPU market. But it also operates in a duopoly in the enterprise GPU market. And it has huge market shares in both! The chart below shows how Nvidia have converted their dominant technology into rapid revenue growth….
Nvidia Nearly Up With Netflix!
Right, so let’s wrap this up and put it all together. Nvidia has clearly got a lot going for it… to put it lightly! And the chart below shows us how the company’s revenues have progressed over the last decade.
Last week we marvelled at Netflix’s impressive 24% revenue CAGR since 2014. Well, Nvidia has managed a similarly impressive revenue growth rate of 23% CAGR since 2014!
2nd May 2023
And that is a wrap! Tomorrow we’ll crack on with looking at Nvidia’s margins and see whether they’ve been able to convert this impressive revenue growth into profits!
Anyway, have a great day team!
The Business Of Team