Greg Lavender wears many hats at Intel. He is executive vice-president, chief technology officer (CTO) and general manager of the Software and Advanced Technology Group (SATG). As CTO, he is responsible for driving Intel’s future technical innovation and research programmes, and as general manager of SATG, he is responsible for defining an artificial intelligence (AI) software stack to support the company’s range of business and hardware offerings.
Lavender began his career in the early 1983 as a network software engineer. After completing his PhD in networking in 1993, he became a professor of computer science at the University of Texas in Austin – a position he held for 14 years. But that was not all he was doing during that period. He was also involved in starting several companies.
“I realised as a professor that I was creating all this useful software, and then other people were monetising it, so I decided that rather than publish papers, I needed to monetise my ideas,” says Lavender.
He found that taking research ideas and making them practical was a lot harder than sitting in an ivory tower and dreaming up new research ideas. But Lavender also discovered that he had a knack for picking the pure research that had practical utility. That knack continues to serve him well today – as CTO, he is now responsible for driving Intel’s future technical innovation and research programmes.
“Because I worked commercially in the early days of the internet, I got used to looking for where the hockey puck was going,” says Lavender. “I created a series of companies. The last one was acquired by Sun Microsystems, where I then spent a decade working.”
“I realised as a professor [at the University of Texas] that I was creating all this useful software and other people were monetising it, so I decided that rather than publish papers, I needed to monetise my ideas”
Greg Lavender, Intel
His last job at Sun was running Solaris engineering. That’s where he got to know Pat Gelsinger, who is now CEO of Intel. At the time, Sun was already into the x86 business, but not with Intel. It worked with AMD Opteron dual-socket/dual-core systems. As CTO of Intel at the time, Gelsinger wanted to do business with Sun.
“Pat gave us some NRE [non-recurring] funding to get the Solaris x86 operating system running on Intel products,” says Lavender. “That’s how Pat and I built a co-engineering partnership. And then he hired me as the CTO of VMware. Then he left VMware to become CEO of Intel. Two years ago, he hired me as the CTO of Intel, despite my attempts to retire.”
Confidential computing and data privacy
One of the challenges with artificial intelligence is that it consumes a lot of private data during the training phase, so Intel is addressing the challenges around data privacy.
“Our confidential computing capability is also a privacy-ensuring capability,” says Lavender. “Europe is ahead in this area, with the notion of sovereign clouds. Intel partners with some of the European governments on sovereign cloud using Intel’s platforms for confidential computing. The privacy-preserving capabilities are built into these platforms, which beyond government, will also be useful in regulated industries like financial services, healthcare and telcos.”
“We also see a convergence in AI that will open up a big market for our privacy-ensuring software and hardware,” says Lavender. “You spend a lot of time prepping your data, tagging your data, getting your data ready for training, usage or inference usage. You want to do that securely in a multi-tenant environment. Our platforms give you the opportunity to do your training securely between the CPU and the GPU, and then you can deploy it securely in the cloud or at the edge.”
“I’m talking with a lot of CIOs about this technology, because data is now such a valuable thing. It’s what you use to train your models. You don’t want somebody else to get access to that data because then they can use it to train their models and offer competing services.”
Lavender identifies edge computing as another area where confidential computing is required. “The edge where all the excitement is going to happen, because after you train everything, all the inferencing is going to happen there whether on your smartphone, your watch, or your laptop. That’s where the explosion is going to be over the next 10 years,” he says. “But AI models that run at the edge are more vulnerable to hacking. So we see that as a market for our confidential computing technology.”
Investing in Europe
Intel intends to invest up to €80bn in Europe over the next decade, funding activities from R&D to manufacturing and state-of-the-art packaging technologies.
In Germany, Intel announced plans to build a $33bn chip manufacturing site, after the German government pledged to cover one third of the cost. Two facilities will be set up, the first of which is expected to start production in four or five years. Germany’s vice-chancellor, Robert Habeck, said in a statement that this would be the biggest investment ever made by a foreign company in Germany.
In Spain, meanwhile, Intel has begun a new phase in its partnership with Barcelona Supercomputing Center. The partners have been collaborating on exascale architectures over the past decade and now they are establishing joint labs to develop a zettascale architecture for the next decade.
“I was in Europe this summer, meeting with a European government on sovereign cloud,” says Lavender. “There’s a lot of ambition, but the money from the Chips Act hasn’t yet been released at a state level. Some things have slowed down, partly because of geopolitics. But we’ve committed ourselves to Europe. And I’m looking forward to working more with our partners there.”