One Does Not Simply Walk Into the Data Center

Earlier this week, the excellent Anandtech posted an interview with Alex Katouzian, the general manager of Qualcomm’s compute products including Snapdragon, Nuvia and a bunch of other important chips. This article sparked strong feelings with us both positive and negative. We are going to dig into it here because we think it sheds a lot of light on how Qualcomm is thinking strategically.

Katouzian is a long-time Qualcomm manager. He came to the company through an acquisition 20 years ago, an acquisition that has given the company some of its best managers. So he speaks with not only the credibility of his very senior role, but also as someone who knows all the history of the business, first-hand.

The big theme that leapt out for us from the article was the notion that Qualcomm now thinks of itself as a general compute company, not a wireless company that has general purpose compute, but a processor company whose largest market is wireless. For those who have known the company for a long time this is fairly radical. They have been talking along these lines for a few years now, but have never exactly come out and said it this way. Time and again in this article, Katouzian talks about some market that Qualcomm is going after with its processors – mobile first, but many times it sounds like they see it as PC first, then automotive and all the rest. This is an important shift for the company, and likely to prove a sensible move, especially given that the smartphone market has reached steady state, low-single-digit growth.

Another theme is the importance of recently-acquired Nuvia. The Nuvia team, which has a stellar pedigree, is clearly being tasked with a heavy load. They seem to be designing the central processor architecture which Qualcomm then hopes to repurpose for various end markets.

This makes sense especially in the context of how they view the future of processors – heterogeneous compute. The best way to think about this is to look at Apple’s M1 processor. This is not just a CPU, it includes a GPU, multiple AI accelerators and a hots of other functions. And this is the heart of Qualcomm’s strategic smarts. As we put processors in more things that do not look like computers or phones, they are going to need to handle a lot of varied functions. This combined with the decline in Moore’s Law means that the chip formerly known as the CPU is going to start coming in many different flavors, and Qualcomm actually seems really well positioned for that.

You can stop reading now, those are the main points of the interview, and they are very important points.

That being said, we have some problems with many of the details.

First, despite all the talk about computing and processors, Katouzian uses the word “Software” only once, and that is in reference to Microsoft, as in ‘we will do the chips and Microsoft will do the software’. This strikes us a glaring omission. Running software is the whole point of chips, especially general purpose processors. The software for mobiles is a very closed ecosystem. Qualcomm essentially wrote most of it for the industry. Organizationally they know this, they probably have 5,000-10,000 engineers doing some form of software work. But strategically it seems that they underestimate the importance and difficulty of doing software for other markets.

Nowhere is that more obvious than in the numerous hints Katouzian drops about winning business in the data center. This quote in particular raised some hackles:

That’s exactly what’s happening with the Nuvia team as well. And I would say, on the server side, that those are opportunistic businesses that we’ll go after. So we’ll see what happens.

The implication here is that Nuvia, which was building a server CPU prior to the acquisition, can build a CPU for PCs, an apps processor for smartphones, and then just walk into the data center. Data center requires a lot of software competence, and there is no sign that Qualcomm has that nor any interest in making the significant investment required. Katouzian probably means they have a narrow opportunity for a limited product type, such as an AI accelerator where the data center owner does all the software work internally, but even this may prove a stretch.

Another problem with their unspoken new strategy is that it leaves open the question as to what happens to Qualcomm’s mobile customers. As we have pointed out repeatedly, Qualcomm’s applications processors are lagging at least a year behind Apple’s. This is creating real problems for Qualcomm’s customers who are now having to go out and design their own chips to compete with Apple. If we were Qualcomm smartphone-making customer we would not be encouraged by some of the passages in this article.

For instance, Katouzian says the first Nuvia product will appear in a PC some time next year. PCs make up a tiny portion of Qualcomm’s revenue, so deploying Nuvia here strikes us as a pretty clear statement of Qualcomm’s intent. And when asked which product Qualcomm is benchmarking against the answer is a resounding “Apple M1”, while we would have thought Apple A15 would have been top of mind. Qualcomm is focused on the PC product, not the mobile. In fact when asked about this gap, Katouzian slaloms around the question, with an answer that basically says their customers have no other choice but Qualcomm.

In all fairness, this is a single interview, with someone coming from a very specific frame of reference (i.e. CPUs and processor). The modem team leader would probably tell a very different story. And as we have said, this is a smart long-term strategy for Qualcomm to move beyond its core, saturated market. That said, we continue to puzzle over Qualcomm’s apparent lack of urgency around the gap between its mobile products and that of Apple’s, and increasingly of its customers. One does not just walk into the data center, but nor should they just slouch out of mobile.

Leave a Reply