Nvdia has announced that they will be buying Mellanox for $6.9bn in an effort to bolster their reach into the datacenter market. That prompted us to think about what has happened to Nvdia over the last few years that led them up to this acquisition.
Technology hardware is notoriously volatile and cyclical. When demand for products is high, companies cannot produce enough product and prices surge. This results in rapidly expanding margins and everything is going well. This is when these companies then start to increase production to meet the needs of their customers. At the same time, as customers face a shortage of product, they often order more than they need (known as double ordering) and inflate the true demand of the product. At some point, demand dries up, and suppliers are left with excess capacity and inventory. This then causes prices to decline and margins quickly recede. Occasionally, investors agree that a company is no longer subject to these cyclical forces and make the claim that “this time is different”. Often, they are wrong.
Nvdia was a company that has rapidly changed their end market exposure. They make the best GPUs in the world, which were widely used by gamers and video editing professionals. These chips now have found new applications. They are used in artificial intelligence and autonomous driving. Being the only major supplier of GPUs, Nvdia became the hardware company supporting the future of technology. Serendipitously, it also turns out that GPUs were the best hardware to mine bitcoins, and as we dove headfirst into bitcoin mania in 2017, demand for Nvdia graphics cards found yet another leg of support. For the company, it was the dream of a lifetime. In a span of a few years, Nvdia has successfully diversified growth into multiple areas of tremendous growth, and there was no close competitor in sight.
And then things started to turn. Bitcoin prices started to fizzle which makes it much less profitable to buy expensive Nvdia graphics cards to mine them. Tesla also announces that they will be developing their own chips for autonomous driving while major cloud players (Amazon, Microsoft, Facebook) all announced that they are developing on their own AI chips. Google has already been using their own AI chips in mass production. The growth trajectory of Nvdia is not so clear anymore.
First let’s point out 2 assumptions I’m going to make:
1. If you think a certain asset is the future of your business, you should own it and not rely on someone else.
Simply put, if you are a multi-billion tech company who firmly believes that AI will drive the next leg of growth, you want to own every part of that technology. If you want strict control over your own diet, you want to cook your own meals. Similarly, while Google and other internet giants might be willing to buy some third-party AI hardware today, eventually they will want to make their own.
2. If you’re going to be buying a lot of something that is demanded by a lot of your competitors, build your own supply chain for it. (Especially true if your competitors all have billion-dollar budgets)
If your largest competitors are out there buying up a commodity you desperately need, you are either going to pay through the roof for it, or you are just not going to get your hands on the asset (and hence lose out). There is no win. Of course, the smart people at the internet giants (and Tesla) are aware of this and wants to have a diversified supply chain. Nvdia’s success in diversifying their revenue into the cloud giants, also paved the way for a series of worries down the road.
Longer term investors would probably point to how annual data looks fine. Nvdia made $6.9bn in revenue in 2016 and grew that to $9.7bn in 2017. Of that growth, $1.2bn was from gaming and $491m was from datacenters. From 2017 to 2018, the growth from gaming grew (but partly driven by sale of cards to cryptocurrency miners) to $1.5bn while data center growth is now $1.1bn.
However, the quarterly numbers show a different trend. Datacenter is arguably the most important strategic piece of business for Nvdia as they move away from a gaming company into an AI hardware provider. Not only has the growth rate here been decelerating, in the latest quarter, the absolute revenue amount is less than the quarter prior. This is why Nvdia is buying Mellanox. It is an attempt to revitalize their exposure to the high-end datacenter. In order to protect their exposure, Nvdia needed to sell multiple products into different parts of the datacenter (i.e. move from the core to the edge) and Mellanox allows them to do that.
Tracking inventory levels are particularly useful here since it gives us both a sense of gross margins going forward and recent share shift trends. Nvdia has consistently managed their number of inventory days at a level that is below AMD until recently, where we saw their day inventory levels double in 2018. This is commonly due to a lack of demand for their products. When they do finally sell these products to bring down inventory, prices (and therefore gross margins) should decline.
The relative inventory levels between Nvdia and AMD depicts something more worrying. While Nvdia is growing inventory days, AMD is experiencing the opposite. This means that Nvdia is producing more product than they are selling, but AMD is selling more product than they are producing. This often happens when the incremental demand for product is moving away from Nvdia to AMD, which means AMD is taking share from Nvdia.
Just 2 years ago, the growth trajectory of Nvdia was undisputable and there was no close competitor in sight. Today, it seems like none of that is true anymore.
Originally published at www.daloopa.com.
