Qualcomm CEO on Data Center Chips, Adding Meta as a Customer

[00:00:00] Speaker 1: Can we just start on actually the near term, because the $5 billion data center number for fiscal 27 is really interesting to a lot of people. I think they really want to understand the composition of it, right? Data center can mean a lot of things. Where is Qualcomm going to move most quickly? [00:00:15] Cristiano Amon: Yeah, very good. Happy to talk to both of you guys. Look, this exciting time, especially because we kind of upgraded our non-handset revenue as we've been diversifying the company from 22 to 40 and enter the data center market. And $5 billion in fiscal 27, it's a high confidence number that we have based on customer engagements we have right now. Most of the revenue is from custom ASIC engagement that expanded since the acquisition of AlphaWave. And you started to see in the revenue some of this innovative technology we came up for accelerators. It doesn't require HBM memory. It's our high bandwidth compute. And as we keep executing on those customers, we're excited to see that revenue going to about $15 billion in fiscal 29. [00:01:04] Speaker 3: I am curious, Cristiano, when it comes to those longer-term targets and the bridge between the short-term, so what you have put out there for 2027, is there any sense? I know you haven't named the customers that get you over that bridge here, but is that going to fill the hole of that $5 billion based on the $1 billion that we already know is out there? [00:01:23] Cristiano Amon: Look, the $5 billion, as I said, I'll give you the composition. I think what I can say right now, I think, as you'd expect, a lot of those contracts are governed by non-disclosure. We have one large hyperscaler customer in the United States, one large hyperscaler customer in China. And then we have some other engagements that we had on the data center. That is all into the $5 billion. So the $5 billion is a very high-confident forecast to which we have the customers, the capacity, the memory allocation, and everything. As we think about the engagement we have today, and also we talk about a contract that we have with Meta, also for two generations of the CPU, especially as we position ourselves for the agentic CPU opportunities, that those engagements are into the $15 billion number. For Qualcomm, as a new entrant, those are great numbers, but really, when you look at the scale of what's happening at the data center, you would imagine as we execute and we show our differentiation, there could be upside. But we're just, you know, being focused on the engagements we have today. [00:02:33] Speaker 3: I do have to ask, though, and investors seem pretty pleased with what they heard yesterday about you articulating this longer-term data center strategy. But I'm curious. I mean, you famously exited the data center business back in 2017, 2018, and I know there were a lot of direct, a lot of reasons for that that went beyond your operational control. But there was a lot of competition then. There's probably even more competition today. Why do you think you can get it right now compared to what it was eight years ago? [00:02:59] Cristiano Amon: Yeah, look, the circumstances, I think, when we started on the CPU, actually, I know people talk about data centers in general term, but we actually started a long time with an ARM-compatible ISA CPU very early. There was a big software barrier. This is before AI, and, you know, at the time, I did not, you know, the market wasn't ready for it, only until later. So it was kind of the, probably the right decision for Qualcomm at the time. This is a different conversation. It's a different data center right now. This is about AI. And, but it's a valid question, right? So I will give you the same answer that I gave when we entered the automotive business, when we entered the industrial business, and everything that we have been doing to diversify the company. Those are fast-moving markets now, changing with technology. So where we really focus is not what the data center is doing today, but what the data center is going to be doing tomorrow. The data center is now moving to a Gentech. That's why you see CPU companies on the rise. You see demand on the rise. The data center is being disaggregated. It's not one single GPU that does everything. You have different hardware. And we're coming out with a comprehensive portfolio. We have a CPU. We have a new memory solution that is not a memory from the memory provider. It's actually been designed by Qualcomm. It doesn't require HBM. We have a different accelerator for the desegregated compute. And we have our semiconductor scale, which is good for custom ASICs. So that's what we're doing. And, you know, if technology matters and you have fast innovation cycles, there's room for Qualcomm. [00:04:42] Speaker 1: Bloomberg Tech, live on Bloomberg Television and Bloomberg Radio, all around the world. And we're speaking to Qualcomm CEO, Cristiano Amon. And a big part of Qualcomm's story is custom silicon now, ASICs. You mentioned it a moment ago, Cristiano. Bloomberg reported on May 26th that ByteDance was one of those hyperscaler companies. And at the time, Qualcomm didn't comment. How is that going to work? You know, there is a distinction between an ASICs program and export of accelerator, which is your own name on it, within the current confines of the U.S. export controls that are in place. [00:05:18] Cristiano Amon: Well, we're still not commenting on it. But I will tell you that, like everybody else that has a significant portion of their data center revenues, also with customers in China, there's a very clear set of regulations and specifications that you have to follow. And you should expect that Qualcomm is complying with those. And I think there's different types of products that are destined for China. And those are the ones that we are shipping to our China customers. [00:05:50] Speaker 1: Thank you. The data center business is multifaceted. So ASICs, your own accelerators, and now CPU. But the market's in a place where NVIDIA or AMD, on an annual basis, bring a new generation of technology, right? And so the question that a lot of people pose for you, Cristiano, is by the time the accelerators come to market and the CPUs in 2028, and customers deploy them, will they not be obsolete? Is it not too late to kind of get into those markets when you look at the peers and how they deploy? [00:06:23] Cristiano Amon: Absolutely not. Look, I get this question all the time. And I think if that was true, Qualcomm would not diversify from phones and enter the automotive industry, enter the industrial, enter the robotics, enter the broadband. And I actually believe it's exactly what I said before. I'm going to give you a simple answer. There are four vectors of differentiation in the accelerator. First of all, we know that power matters. If you look at the demand that exists in the data center right now, it's a demand that's going like this, and the energy available is like this. So that creates, it's perfect for Qualcomm, because what we do is very power efficient. We design our technology, always thinking there's going to be a battery on the other side, and we're going to have efficient power consumption. That's one vector of differentiation. The other one is what I said. We design for the future, which does not have the memory constraint that you have today. We don't have to move a lot of data between the accelerator to memory, and that reduces thermal, reduces power consumption, and increases the memory bandwidth. And we don't have to use expensive HBM. We have the ability to do this at a much lower cost. Number three is we are coming in with very high density of transistors with our accelerator, and we're not a small company. I think we ship 40 billion components a year. We do more than 30 new chips per year on leading nodes, and I think we have the ability to execute. And if the data center is changing because of Agentec, and the data center is changing because of disaggregation, there's an entry point for Qualcomm. [00:08:00] Speaker 3: I am curious, Cristiano, particularly with Meta and Dragonfly, how confident are you that Meta will still be there in 2028 as a partner? And I ask that only not so much from the Qualcomm side, but Meta has sort of moved around a little bit with some of its strategies and how it's approaching AI. Is the agreement that you have with them, is that basically firm? Look, two parts. [00:08:25] Cristiano Amon: Yes, we have a contract with Meta. That's why we talk about them as a customer for two generations, and it's very significant for our CPU. We have a very unique CPU. If you look at some of the metrics we provided, and I think we have a track record to have a good CPU. The second part of the question, I will never bet against Meta. You know, they have incredible first-party applications. They have a lot of data. They have a lot of data from consumers. They're well-positioned. We are on the early innings of AI. Anybody that believes right now that whoever the players are on AI, this is what is already defining. This is a decade long, and I think if we learn something about the Internet is technology is going to move very fast, and the players are going to change. I will not bet against Meta. [00:09:14] Speaker 1: Cristiano, can I just jump in on Meta really quickly? What I find really interesting about it is the relationship was already there, right? Snapdragon and Quest. I spent a lot of time down in Palo Alto on the Orion project, kind of looking at what's under the hood. How much did that existing relationship on the consumer electronics side get the win on the data center side? [00:09:33] Cristiano Amon: It's 100%. Look, at the end of the day, this is not one-time relationship. We have built over a long period of time a strategic relationship, and what happened was we have those relationships in place, and we have more products. You have the trust. You have the capability of execution. And by the way, this is also true what happened on CustomAzik. We bought Alpha Wave. They have some engagements, but now they're part of Qualcomm. Qualcomm has broader relationship with some of those companies, and I think that just scale, and that's exactly what you're seeing. [00:10:04] Speaker 3: Well, can I ask you a little bit about that? You mentioned Alpha Wave. You've obviously, you have Ventana. You have the modular deal. You made a lot of acquisitions in a relatively short period of time. I guess my question is twofold. A, should we expect more anytime soon? And based on the acquisitions that you've done, have you actually already sort of integrated that into some sort of sellable product that's already sort of out there as one? [00:10:29] Cristiano Amon: Yeah, 100%. Look, if you look at our track record, right? So especially as I think about the diversification of the company, first was automotive. We bought Arriva, and that is now the stack that you see on BMW for highway autopilot. We bought Nuvia, and you see now our CPU showing up, not only in PCs and phones and automotive, but now in the data center. We bought Alpha Wave, and you see us now into the connectivity business and the custom ASIC business. And the modular acquisition is incredibly exciting. I actually don't think the industry investors really understand what that acquisition is. I think my prediction is people are going to read it three or four or five times, and eventually they're going to get it, which is an industry-friendly alternative that we're going to make it available upon close to everybody. So all of those things are- And the modular acquisition, Chris Lattner, he's sticking around? Absolutely. That's why you can see, if anything, if anything, what you should see about this acquisition, by being an all-stock deal, it means that the founders are now fully vested on the future of Qualcomm. And I think there's another thing that investors do not understand. They ask me the question, why is this size acquisition doing all-stock? It's because they're all now committed to the future of Qualcomm when we're connected together. So all of those things is part of us doing that. And to your question is, there are going to be some other acquisitions, but as we have been very disciplined, as we're building industrial, some of the acquisitions we did, Arduino with 33 million developers would focus on industrial, Edge and Pulse. Now you see us doing data center. Future acquisitions are focused on data center and robotics, which is another area of investment for Qualcomm. [00:12:14] Speaker 1: Cristiano, smartphone and AI PC is still there for you, right? And you'll be aware of the world that we woke up to this morning in memory with Apple raising prices as well. Has your outlook for the PC and smartphone markets changed or shifted at all for this year, next year? [00:12:32] Cristiano Amon: That's a great question. I think you have to look at those markets very short term, and you need to look at this market, so what's happening in the midterm. For the very short term, I think the whole world is dealing with this memory situation right now, which is not great. And I respect the companies kind of make choice to different industries, but then what happened is you don't have a demand problem and mobile NPC, you have the memory availability and prices. I think you saw the message from Apple. So we even expect, when we think about, for example, the Android market in China, I think when you think about year over year, when you look at 27, expect it to be flat to probably a down bias. And that has all to do not with demand, but with the memory situation. But the other part's more interesting. There's a lot of clarity right now because of agents, how this market is going to change with AI. Very, very clear. And I think we see a lot of those phones now becoming the customers of the phone as the human and the agents. And that is changing the architecture of the devices.