AI, GPUs, and focus: 7 takeaways from the AMD CEO Lisu Su’s SXSW keynote

AMD has bet big internally on AI in a goal to increase its chip output, chief executive Dr. Lisa Su told an audience at SXSW on Monday. And that includes her, too — as she uses it to manage her calendar and send email.

In a conversation with podcaster Ryan Patel, Su outlined the number of ways in which AI is being used, stating that it would become virtually ubiquitous in a few years. Here are the most important takeaways from her conversation.

“I think the best thing to kind of keep in mind is I think AI is the most important technology that has come sort of on the scene over at least the last 50 years,” Su said.

You can watch the conversation here.

AMD has bet big on AI, internally

AMD ships AI hardware inside accelerators and PC processors, but also uses it to accelerate its own business.

“So at AMD, we want to be at the bleeding edge of using AI through every aspect of our business,” Su said. “We’re using it to design chips, we’re using it to design faster chips, make them more reliable, build better software… it’s a productivity tool. And my goal is I’m telling my engineering team I’d like to like you know, increase the number of products we can get out every year with the use of AI.”

Su said that AI is used within its HR processes, its financial processes, and customer services as well.

It’s not perfect, Su admitted, and one of the ways AMD thinks about “responsible AI” is about training it the right way. “AI that we have to be careful about, its how you train models, making sure that the data that you’re using isn’t biased in some way,” Su said.

“But what I will say though, is the answer of those who are a little bit worried about what AI will do, the answer is not, go slower,” Su added. “Okay, that is definitely not the answer.”

AMD’s GPU supply will increase

In response to a question from the audience — a representative of AMD’s AWS cloud service — Su said that AMD’s GPU supply would increase. With a grin, Su promised to help out.

“In the chip industry, we do go through these cycles where demand is higher than supply,” Su said. “One thing I can say for sure is we’re really good at ramping up supply. So, you will see the supply for GPUs increased significantly. Every quarter, you know, going forward and that’s our job. Our job is to build capacity there.”

Su uses AI as well

“I’m absolutely using [Microsoft] Copilot and it’s one of the things that we use when you think about, you know, very easy things to do, you know, summarize meetings, track action items, you know,” Su added.

What she doesn’t do? Use it to write email. “It doesn’t do it that well,” she said.

Who wins in AI? Local AI or the cloud?

AI consists of two aspects: machine learning (or training) and testing, also known as inferencing. It’s the latter that we think of when we talk about LLMs or AI chatbots, as well as AI art. But it’s still faster to run it in the cloud. It might always be faster to run it in the cloud, but AI will be more pervasive on the client like AI PCs.

“When you think about inference — inference is basically when you ask a question, you get, you get a response, and a lot of that almost all of that happens in the cloud today,” Su said. “And frankly, that’s, that’s some of the latency that I talked about. It takes a few seconds for to answer.

“I do think a significant percentage is going to happen at the edge whether in clients or in places closer, closer to your person, and it’s just going to be a matter of you’re going to see AI and all of these different places,” Su added. “We do think that inference will surpass training as we go through the next year or two. So that’s where we see the AI trends going.”

What did AMD need? Focus

Su is nearing her ten-year anniversary as CEO, which takes place in October. She looked back at her earlier career, and said that the thing that AMD needed was focus.

“When I first took over as CEO, like any company that doesn’t have quite enough resources, we were probably doing too many things,” Su said. “And it was very important for us to decide what we were going to be good at. And what we’re good at is building the highest performing computing capability out there. So we’re a high performance processing company. That’s what we spend our time on. And it turns out that that was not a bad bet.”

When asked to describe herself in a word, Su replied that she called herself a “doer.” “And our goal, my goal is to set very ambitious goals and then help the team get it done,” Su said. “So that’s why I call myself a doer. I love going into the labs. I love seeing what’s the latest and greatest going on and I love spending time with our customers and partners.”

Career memories that stand out

Since Su has worked both as an engineer (with undergrad, masters and a doctorate from MIT) and as a manager, she said she had lessons from both.

“If you’re an engineer, I can say for sure [the lesson learned was from] the first very, very first product I worked on,” Su said. “I was working at at IBM at the time, we announced the product that day and I can tell you nothing was working. Like the chip wouldn’t turn on. Those are scary times when that happens.”

Su said engineers are “super innovative and creative, and when there’s a will, there’s a way.”

As far as a business lesson, it was the “the very first business I managed,” Su said.

“I was like, I was an engineer turned business person,” Su said. “And no offense, but I thought it was going to be relatively easy. I didn’t think there was going to be a big transition between engineering and business. So for the very first business I managed, I literally missed my first annual operating plan by like 50 percent. I was devastated.

“My boss told me, Lisa, You’re terrible. You’re awful. How could you do that? And I said, You know what? He was right,” Su added. “I could have done better and so I’ve learned from that I learned I learned how to ask more questions next time.”

What’s Lisa Su’s workout routine?

It was an odd, personal question, but Su graciously answered it. “I have a great person who comes to my house every day or every few days anyways, and it’s it’s boxing and strength training,” she said.

When asked whether she prefers F1 racing or golf, Su said she prefers the latter.

CPUs and Processors