Roughly a year after Intel installed CEO Pat Gelsinger to lead the company’s turnaround, it still has a lot to prove.
Its earnings report last week demonstrated signs of life, yet the company still has a long way to go after a prolonged period of manufacturing weakness. Gelsinger has vowed to regain Intel’s manufacturing leadership, and part of the strategy involves massive capital investments in new factories in Ohio, Arizona and Europe that are taking a toll on profits.
Another equally important task Gelsinger embarked on as part of his effort to reinvigorate the Silicon Valley giant was remaking the company’s executive leadership. Over the summer last year, for example, Intel jettisoned its data center boss, Navin Shenoy, and split the group into two parts. At the same time, Intel named former VMware executive Greg Lavender to the CTO job.
A computer scientist by training, Lavender's career in tech has included a teaching stint at the University of Texas at Austin, a decade at Sun Microsystems and running cloud infrastructure for Citi; most recently, he served as CTO of VMware while Gelsinger was CEO there. In a recent conversation with Protocol, Lavender talked about how he and Gelsinger are redefining the CTO role, the strategy behind Intel’s software efforts and how he approaches managing the thousands of technologists he is responsible for leading.
This interview has been edited and condensed for clarity.
Can you tell us a bit about your role as the CTO of a semiconductor company?
As the CTO, you get pulled in a lot of directions. My No. 1 priority is building my leadership team, and I’ve made some changes there and continue to recruit both internally and externally. I like to tell people that I know a little about a lot, or a lot about a little. I balance my time between chief talking officer — which includes doing things like this — and chief technology officer, which is the tech side. I sit on Pat’s staff, and we have a software-first strategy, and I’m co-chair of the AI/ML strategy for the company.
Pat gave me two jobs. I’ve got Intel Labs and a bunch of advanced technology development going on. I created a security group pulling together 1,000 security architects, firmware engineers — systems people — to take a hard, hard run at confidential computing. And then I collected another 5,000 software engineers across the company into the Software Advanced Technology Group, which is everything from BIOS to firmware to compilers, operating systems.
I could spend the rest of the call telling you about all the open-source code that we put out there that most people probably don’t know that we did because the past administration didn’t want people talking about these sorts of things. The world is waking up to our open-source code contributions that are already there.
Why did Pat ask you to join Intel? And why did you take the job?
Pat just said, “You’re the right person, at the right time, for this job.” I had to let that sink in a bit. I have great respect for Pat, I’ve known him for 15 years, since I was at Sun and he was CTO of Intel. Think about it: The first CTO of Intel asked me to be his CTO. It’s a real privilege, I felt very flattered by the opportunity. He and I have not failed yet, and we’re not going to fail at Intel. We’re going to make it great.
What’s your plan to help with Intel’s turnaround? What can the CTO do?
Let me give you some context. The Intel Developer Forum was the premier technical geek-out event. It was the place that all of the coolest technology from Intel, and all of their customers, would end up. And it’s like, Intel somehow lost that mojo. [Editor's note: Intel Developer Forum was shut down in 2017.] So for me, when I came into the door, I wondered: How do we get that mojo back?
The issue is that what we’re doing hasn’t been communicated to the community, and the community, therefore, wasn’t paying attention to all the great things going on. So now, under Pat, we’re much more vocal about all the things we’re doing, particularly in the software area.
I’m trying to create a pull function for software. Right now we push a bunch of software: “Here it is, here it is, hope you use it.” But you’ve got to be in the ecosystems, because for low-code/no-code developers it’s about productivity. They assume the performance will be there by the cloud vendor, or whoever is writing the infrastructure.
So how do you get the software out into the market, in the hands of developers, either without them even knowing it — like putting it into PyTorch or TensorFlow or these other ecosystems — and make it so ubiquitously available in the ecosystem that you’re getting all the benefits? It’s open source or freeware, and it's available to you without having to perform any kind of brain surgery. And you can just get all the benefits on our platforms by picking up these ecosystems.
As a relatively new person at Intel, where staffers can stick around for 10, 15, 20 years, how do you think about running the CTO’s office? Do you have a chief of staff, for example?
I’ve never liked the term chief of staff, because I don’t want some police officer guarding my door. I like to be accessible, and in fact, my staff thinks I’m a little bit too accessible. As a new leader, it’s important to me that people understand who I am, how I communicate. Recently, my big teams had their own internal offsite sync-up meeting, and I came and spoke to them for the first 45 minutes, just to confirm the strategy and how to execute it.
When I worked in banking, I never adapted to the hierarchy [Editor's note: Lavender worked at Citi for six years]. Look: Andy Grove is famous for having the flat organization, that he pioneered that philosophy and theory in Silicon Valley. I’m very used to egalitarian environments.
I want to spend at least 65% of my day, if not more, with the technologists. And the other 25% to 30% of my day, I’m dealing with finance and lawyers. I mean, that’s just part of modern business, right? My philosophy is that if the “tax” on my time is more than 25%, then I’m going to put more structures in place in order to deal with it so it consumes less.
As a computer scientist and not a chip engineer by background, how do you tackle some of the decision-making around the technical complexities of semiconductors?
I’m not a [hardware] engineer by background. I find it fascinating, the material science level. I studied physics in undergrad, so I sort of had to dust off my old textbooks, get the latest editions, to go read more on semiconductor physics. I ask all the right questions because I have a lot of experience. As a professor for 14 years at the University of Texas at Austin, I supervised lots of undergrads, master’s and Ph.D. students, so I’m big on educating people.
When you present some of the more technically complex aspects of the groups you’re responsible for to the board of directors, how do you go about calibrating your explanation? What is different about talking to the board versus someone like Pat?
We send a short white paper that has all the definitions of all the acronyms, so you sort of give them a decoder ring because there are a huge number of acronyms when we start talking about AI or ML, or PyTorch/TensorFlow. It’s just a plethora of technologies — some of the board members are clearly technical. But we always try and give them a pre-read, so they’re briefed. Then they can ask questions before the board.
A lot of what we discuss is confidential, but they’re very engaged, and ask a lot of great questions. But the board is always concerned about the competition, and we feel like we’re going to show up better. My philosophy is that honesty and transparency are the things that you should adhere to in a board meeting: being open about where we are, and what our plans are, and taking on all the questions.