《财富》专访苏姿丰：我怎样再造AMD

In this in-depth interview, the AMD chief talks supercomputing, executing a turnaround, and running a global company in the midst of a pandemic.

This edited Q&A has been condensed for space and clarity.

ENGINEERING A NEW AMD

When you took the reins in October 2014, AMD’s revenue was down almost 40%; your market share had been cut in half—your stock even dropped below $2. But your strategy for revamping the business by focusing on higher-performance chips has been a success. Your stock is up,1 and reviewers have gone crazy over your latest Ryzen chip lineup; this year 100 new laptops are coming to market with AMD chips. You once described the turnaround process to me as “fighting your set of wars.” How does it feel to have won some battles?

SU: It’s been very exciting, rewarding—all of those things. The past couple of years have certainly helped build that confidence that, hey, when we set out to do something, we can actually get it done. And when we began, there was a lot of convincing to do. But there’s a lot more to do, and in our world it’s always about what’s next and what’s the next big thing.

Now the expectations are high. It’s interesting, because life, as well as product road maps, is all about making choices. And the choices don’t actually get easier as you do better. The choices are exactly the same, if not harder, frankly.

Your product road map over the past five years included whole new designs for CPUs and GPUs.2 GPUs used to be just for video gaming, but now they’re also being used in data centers to help compute big-data analysis. What comes next?

We’re making large investments in graphics and what we’re doing around optimizing graphics for both gaming as well as computing. That’s a new vector where we’re putting a lot of emphasis.

If you think about what differentiates AMD, it’s the idea that we can put the best processors together for each of the workloads. This idea of bringing CPUs and GPUs together in different combinations and with different interconnections really goes toward where accelerated computing is going in the future.

It’s kind of funny that the same kinds of GPU chips that video gamers needed also turned out to be the thing that’s running A.I. and machine-learning apps in cloud data centers at Google and Amazon. Why are those kinds of GPU chips, which can run lots of simple tasks very quickly, so in demand in data centers?

It’s the idea that computers can get smarter and smarter. And the way they get smarter is they get better at recognizing patterns and matching patterns and then using that information to become a little bit smarter in the future. And that’s the whole concept of machine-learning and artificial intelligence and high-performance computing.

This part of computing is actually moving faster than anything else because we have this tremendous amount of data that we’re generating, which we don’t really know what to do with. Each of us is generating so much information. Our companies are generating a ton of information. The Internet is generating a ton of information, and we need to figure out what to do with it all.³

How do you bend the performance curve? In technology, if you plot the performance gains made by our industry over a five-or 10-year period, it often looks like a straight line. You can draw a straight line through it, and our goal in life is to change that line. We want to be above the line, bending the curve.

All about the exaflop

AMD just won two government bids to build some of the fastest supercomputers ever.⁴ How is your technology being used in that context?

If you think about the problems that you’re solving in science at the Oak Ridge National Laboratory, which does medical science and weather science, or Lawrence Livermore, which does more of the national security–type things around simulating our nuclear stockpile, all of them do better when you can take larger data sets and do many more calculations.

That’s what we’re doing in building these large supercomputers. But we’re going above what normal scaling would allow you to do by adding this combination of CPUs and GPUs and high-performance interconnect. And that’s kind of fun because you can see that, yes, it’s the same technology that goes into game consoles—albeit much, much, much bigger. But it’s used in a different way, and it’s used in really tough applications that will change the world going forward.

The supercomputers you’re building will reach exaflop speeds,⁵ five times as fast as the fastest current supercomputer. How can you make such a huge leap forward?

It’s not any one thing in particular, Aaron. I think it’s a combination of things. But the most important is the idea that these components are smarter because they have smart interconnects that allow them to share data and share operations much more efficiently than what has been done in the past.

My new laptop isn’t five times as fast as my old laptop, not even close. Do you envision that consumer devices are going to see huge leaps like that again?

The technology that we’re putting into supercomputers today will absolutely show up in consumer devices. It might take five more years for that to be the case, but it’s always been the case that you use these big applications to drive the barriers of innovation.

Let’s solve the big problems, then, over the next five to 10 years, you trickle that to consumers once the cost point gets there and once the manufacturing technology gets there.

Your typical word processor probably doesn’t need a much faster CPU. But there are some applications that I do think are going to hit consumers. A lot of this machine-learning technology, for example, is really useful in things like speech recognition, right? And if you think about your speech-to-type conversion right now, it’s okay, but it’s still not that good.

Part of your strategy for reviving the company was to get into the business of making custom chips for gaming consoles. Now I’m seeing all kinds of cloud gaming services everywhere, no special device needed. Is console gaming still a good business?

Gaming is a great business. I think the last number, there were over 2 billion gamers if you look at from mobile to PC to console to cloud.⁶ This is a big year for gaming, with both Microsoft and Sony launching their next-generation consoles. They are some of the most anticipated consumer products of 2020. And again, we like gaming because it uses technology very, very well. And we’re able to reach a lot of households, and it will continue to be an important part of our portfolio. I do think cloud gaming has opportunities, but it’s still many years out.

The new normal

The world is now facing a crisis from the coronavirus pandemic. On March 5, you told analysts that the outbreak was having a modest impact on your financials so far. But given your global supply chain, what are your longer-term worries?

The COVID-19 crisis is truly unprecedented and touches all of us. Our priority is protecting the health and safety of our employees, partners, and communities. It’s amazing to see a company of more than 10,000 people transition to work from home on a dime. We’ve also figured out how to do some things differently, including some very sophisticated engineering work remotely. At the same time, we’re supporting our customers as their priorities change. We have a complex supply chain where our products go through multiple countries to get manufactured. Although there were some early disruptions, we’ve been able to navigate it.

Does that mean having more redundancy geographically?

That’s exactly right. It’s having redundancy in your supply chain. It’s having redundancy in your engineering teams. It’s building the notion of, hey, you have your contingency plans as things change. And, in some sense, it’s building a company that can withstand lots of different things related to the environment we’re operating in.

You’re one of just 35 female CEOs in the Fortune 500 right now. What do we need to do to have more women leaders in tech?

One piece is about just the pipeline and having enough people start in the field. And then the other piece is making sure that women have good opportunities. Give good people good opportunities—they will shine.

We are definitely very focused on ensuring that as we look at leadership, particularly in the technical ranks.⁷ That being said, these roles are very competitive, and at the end of the day it’s always about, Let’s get the best person in the job. 

****

Between the lines

1. AMD stock performance

Oct. 8, 2014: $3.28

April 9, 2020: $48.38

Source: Bloomberg

2. Know your chips

The CPU, or central processing unit, is commonly used as the main computing chip in PCs and servers. The GPU, or graphics processing unit, started out helping speed up video games but is also used for A.I. and big-data apps now too.

3. Global Internet users

2001: 495 million (8% of total population)

2019: 4.1 billion (54% of total population)

Source: ITU

4. Super deals

In March, the DOE’s Lawrence Livermore National Laboratory picked AMD to supply processors for El Capitan, its $600 million supercomputer. In 2019, AMD won a similar deal to supply a new supercomputer called Frontier for the DOE’s Oak Ridge National Laboratory.

5. Big, big numbers

An exaflop requires computing 1 quintillion floating point calculations per second—or a 1 followed by 18 zeros. Apple says the A13 processor in the iPhone 11 can reach one teraflop, so it would take 1 million iPhones to equal an exaflop.

6. Play on, players

An estimated 2.5 billion people played video games last year, spending $152 billion, says research firm Newzoo. About 45% of the spending is on mobile games, about one-third on consoles, and the rest on PC gaming.

7. Still a long way to go

Women held 24% of all jobs and 18% of engineering jobs at AMD in 2018, according to the most recent data available. For context: Women held 26% of computer and math-related jobs nationwide last year.

A version of this article appears in the May 2020 issue of Fortune with the headline “The Conversation: Lisa Su.”