How a Tech Company from the 60s is taking on AI, IoT
When we talk about the next big thing in tech, it’s easy to overlook the fact that many of it relies on ideas that have existed for decades.
Take for example artificial intelligence (AI), which was formally founded in 1956 at a conference in Dartmouth College, in Hanover, NH; or the Internet of Things (IoT), which was born in the 1990s. These technologies have spurred billions of dollars of investment and the creation of hundreds of companies. But one of the companies poised for significant growth in these areas is more than 50-years-old itself.
Intel, the microprocessor company founded in 1968, is seeing tremendous opportunity in these emerging technology areas as demand for PCs and chips for traditional data centers soften. Intel’s IoT business was its fastest growing business segment of 2016, with 15 percent growth year over year, while its fourth-quarter sales of server chips to cloud service providers jumped 30 percent year over year.
Partnerships with cloud service providers, the biggest public cloud firms in the world, are helping drive demand for Intel products, Raejeanne Skillern, VP of Intel’s Data Center Group and general manager of its Cloud Service Provider Group says.
“They can consume technology as fast as we can deliver it,” Skillern says about Intel’s cloud service provider customers, which include Amazon Web Services (AWS) and Microsoft. It’s a different pace from 8 or 9 years ago when she first joined the data center group. At that time, “Amazon was just coming on board, Google was growing in search, and Intel’s position was very small, we had about 35 percent market share because AMD was largely winning in that space,” she says.
Intel calls its largest cloud service provider customers ‘Super 7’ – AWS, Google, Microsoft, and Facebook in the U.S., and in China, Tencent, Alibaba, and Baidu.
“Our strategy is fundamentally based on having deep and first-hand knowledge and collaboration with our largest customers,” Skillern says. “They’re kind of unique in the way they grow. We’ve been fortunate enough that they’ve allowed us to go in and have these very direct engagements; I think that’s kind of surprising to some because we’re a silicon ingredient supplier of an infrastructure system – why would Intel as one component have these direct relationships? But it’s because of the deep customization that we do.”
The results of this customization include Amazon’s C5 instances which are set to launch in early 2017 and are based on Intel’s next-generation Xeon cloud processors, code-named Skylake. Amazon says the C5 instances are “ideal for compute-intensive workloads like ad serving, scientific modeling, 3D rendering, cluster computing, machine learning inference, and distributed analytics.”
“For us as we bring the Skylake platform to market obviously getting our best technology into the service provider segment is going to be critical throughout the year,” Skillern says.
Part of this focus is the early ship program where cloud providers can get a hold of Intel technology a couple months ahead of the market. “This is because they’re not as constrained by the OEM ecosystem and the validation schedules that apply to enterprise,” she says.
Over the next 12 months Skillern and her team will focus on driving faster adoption in the cloud service provider segment and helping these customers grow through different services that use Intel technology, including high performance compute as a service, AI as a service, big data analytics as a service and security as a service.
“A lot of people know us as a compute provider but a lot of these challenges are happening at their platform level and software level. When they create solutions they think of the entire stack not a bunch of little, individual ingredient parts.”
“A lot of what we do is we match our engineering teams with theirs to meet at the level that they’re at; whether its platform, hardware, OS,” she says. “As they design their next generation systems we help co-design a lot of the boards and systems with them.”
Aside from Skylake, Intel also plans to launch the next processor in the Xeon Phi family, code-named Knights Mill, to power artificial intelligence.
“Artificial intelligence has been around for decades,” Skillern says. “We’re seeing it becoming very pervasive; talk-to-text, photo recognition and tools, ad matching, all of those analytics capabilities and tools come together to provide better services and answers to important questions.”
Intel set a lot of groundwork in 2016 for its AI strategy, making several key acquisitions, including a company called Nervana, which has a software and hardware stack for deep learning. In November Intel said it plans to integrate Nervana’s technology into a new product code-named Knights Crest, and the formation of the Intel Nervana AI board to further AI research and strategy.
Its partnership with Google will also help to drive its AI efforts, including a focus on optimization of the open source TensorFlow library to provide software developers a machine learning library to drive the next wave of AI innovation, Skillern wrote in a blog post last year.
“Artificial intelligence can bring amazing outcomes in medical science, retail, and business operational efficiency. I also get excited working with it because you’re bringing together such amazing minds. The company we acquired, Nervana, is a talented group of about 50 people who are some of the smartest you’ll meet in the industry,” Skillern says.
“It’s just really fun when there’s something new, and you get to learn it and see the possibilities and impact,” she says.