저자: LisaSu · 외국인
요약 보기
AMD 2025년 실적 호조, 데이터센터·CPU 사업 강세 강조. 1분기는 계절적 비수기이나 데이터센터 매출은 Q4→Q1 증가. 하반기 MI450 시리즈와 Helios 랙으로 변곡점 예상. 최근 60~90일간 AI 수요 가속화 체감, 컴퓨트 부족 지속. 소프트웨어 기업 위협론에 대해서는 AI 툴을 자사 개발에 통합하는 쪽이 승자라고 주장.
다듬은 글
shares of AMD, I I don't even want to say taking a hit because the reason why is the stock's up very big in the last few weeks, but it's coming off today. The company beat the estimates. They boosted boosted by the exactly what Dave was talking about, which is the booming data center business, but also and I think we're really going to miss the point and the analysts missed the point badly by the CPU business, which is the business that competes against Intel.
We got to learn about both of them because the CPU business was completely left out by the analysts. It was kind of pathetic. And that's why I'm glad to have AMD chair and CEO Lisa Sue on the show. Lisa, I'm going to offer you a congratulations because not only are you well ahead of where you told me you would be with GPUs at this time two years ago, but you're also killing it when it comes to CPUs and I think we got to talk both.
Where are you making the most money now and where will you make the most money next year at this time? Well, first of all, Jim, uh, great to be on the show with you guys. Thank you. Uh look, 2025 was a fantastic year for AMD. We really closed it off strong and what we see is strength across the entire business which is really exciting.
I mean you know obviously there's a lot of about you know data center and AI and what we're doing around GPUs and that business has performed extremely well and we're going into a big inflection year here in 2026. But as you were saying like the CPU business is absolutely on fire and this is one of those stories where you know I've always said Jim that you need all kinds of compute when you think about data center and the massive buildouts that we're seeing and we're seeing just continued strength in the CPU business which is you know really our stronghold right now.
So we're very excited about um you know what we've done but much more excited about what we have going forward in 2026. Right. Well, let's take off the table the idea that your guidance disappointed. Your guidance was actually pretty consistent with what you've been saying. Second half, I don't know why anyone would think it was going to be second quarter.
Uh, am I missing something? Did you tell people that it was going to be second quarter and it turns out to be second half and they're disappointed? Where is the disappointment come in? Well, let me say um something, Jim. we think about how the compute market is evolving. Let me just give you a a picture of where the trends are right now in AI.
I know there's a lot of conversation. You know, every morning on the show you guys are talking about where is the AI trade. Um what I would tell you from someone on the inside is AI is accelerating at a pace that I would not have imagined. So you see, you know, AI demand from the hyperscalers, you see AI demand from enterprises and what is the foundation of all of that is compute.
And so what is really important for us is to ensure that you know we have the full range of compute now as we go into the first quarter. The first quarter is usually seasonally down for us uh because we do have some consumer exposure and I think the most important thing is we're actually seeing our data center business um actually increase um from Q4 uh into Q1 which is a show of the strength of the business both on the CPU side which is uh you know going like gang busters and with our AI uh capabilities and GPUs uh we're seeing very nice ramp into the first half of the year but yes we're going to have an inflection point as we go into the second half of the here with our new, you know, MI450 series and our Helios uh rack systems.
Let's talk about that, Lisa, because one of the things that I think is important to people to understand is that yes, it's absolutely true. The Ver Rubin certainly has some qualities. This is the next generation to be unveiled at GTC uh not that far from now. Ver's got car really some terrific stuff, but your stuff when it comes to memory is actually bigger and better.
That that's absolutely right, Jim. What we're seeing today is because there's so many AI workloads out there, uh, you really have to optimize and that's optimized for training and optimized for inference. And from our standpoint, we're just seeing an explosion of inference as enterprises are using more and more AI.
And that's exactly the sweet spot for our MI450 series and Helio. So, we're super excited about how that's coming up. you know, the development is going extremely well and we're working closely with our customers and the entire supply chain uh to bring that uh to market in the second half of the year.
Lisa, I want to just go back to a comment you just made where you said AI is accelerating at a scale I would never have imagined or not have imagined and I just want to understand that a little more. Not have imagined when even recently and what do you mean when you say it and you know again this rate of change is so quickly is it only going to increase?
Well, David, what I would say is, you know, with every period of time, like if you had asked me six months ago, I would say, "Hey, demand is strong and we're continuing to see good signals um from, you know, our top customers." I think when I say that we're seeing an acceleration, I would say in the last uh 60 to 90 days, we have actually seen an acceleration, particularly when we think about 2026.
And you know that comes in terms of you know demand patterns, forecasts, and just the general feeling like there's not enough compute out there for everything that wants to be done. And you've heard that from a number of our top customers, but I can say that we see that on the inside. There's a desire to ramp compute more uh quickly.
And um as we said at the top, it is um not only for the uh GPUs, but we're seeing very very strong demand for CPUs as a result of all of this um AI uh enterprise work that's going on there is actually pulling more CPU content, which is, you know, a great place for us to be. No, and we've watched your stock price respond uh to that as well.
Coming back to the rate of change, we've been talking, you probably listened at the top of the show and certainly after yesterday and many days lately about um software companies very much concerned perhaps amongst investors about their future business given the tools that we're seeing happen almost every day.
You know, how do you view it as somebody who's been in this ecosystem for so many years? Lisa, should investors be concerned about software company's ability to compete when Claude or Open AAI can come up with a new tool tomorrow? I I think the real key is what we're seeing with all of these tools and we're working very closely with the model builders is that um we have to integrate what the AI tools are capable of with our basic uh development capability.
And that's true for hardware guys, that's true for software guys. uh we have to take the power of the tools and actually you know make our products u you know better get them to market faster uh really you know accelerate the rate and pace of innovation and that's exactly what we're doing I think the disruption comes from you know those who are able to take and harness uh the power of AI and the closer that we work between you know sort of the the largest foundational model guys and you know sort of the innovation that we're trying to drive um at the engineering level I think the better the overall product outcome is going to be and you The benefit of that is frankly um you know the entire ecosystem because we're going to accelerate the rate and pace of innovation.
Comments 0