The global technology sector is on the brink of an explosion of data, according to top executives at Intel, who say the growth of cloud computing will be accelerated by new technologies like artificial intelligence (AI), the Internet of Things, virtual reality, drones, robots and autonomous vehicles.
“This explosion of data will usher in an era of profound transformation,” said Murthy Renduchintala, President of Intel’s IoT and Client Businesses. “It will create a need for new infrastructure and drive demand for a new interconnectedness of everything.”
At the heart of this revolution is the data center. At its Analyst Day Thursday, Intel outlined a “Data Center First” strategy that looks beyond the CPU to a product portfolio that spans the data center, including memory, networking and emerging technologies like the IoT and AI.
“We’re shifting from a server and CPU company to a data center company that builds high performance racks,” said Intel CEO Brian Krzanich. “We believe we are now architecting at the rack and data center level.”
That shift is reflected in Intel’s decision to prioritize its data center business over its PC sales, meaning data center customers will now get first crack at Intel’s newest chip technology.
It’s a symbolic shift, given Intel’s history in the PC market. But it reflects the rise of the cloud, with IT workloads shifting to huge data centers that deliver apps and services to a growing universe of distributed devices. Intel is positioning its business to address this trend, which will continue to reshape the computing landscape for years to come.
Cloud Growth Driven by Machine Data
The reaction to Intel’s Analyst Day largely depends on one’s horizon. Stock analysts and investors were focused on the state of Intel’s enterprise data center business, where sales and profit margins remained under pressure as U.S.corporations shift spending from on-premises facilities to cloud platforms – a trend that has clearly gained momentum over the past year. Shares of Intel sold off by about 2 percent during the presentation.
In acknowledging the cloud shift, Intel pointed to a future in which new technologies expand the market for its products. The vision outlined by Intel aligns with our central thesis here at Data Center Frontier – that new technologies will drive strong demand for data center capacity for the foreseeable future. As the global leader in the semiconductor market, Intel is uniquely positioned to assess IT trends.
According to Krzanich, it all comes back to data. He presented a projection of daily data creation in 2020 from various sources:
- Average Internet User: 1.5 GB
- Autonomous Vehicle: 4 TB
- Connected Plane: 5 TB
- Smart factory: 1 PB (petabyte)
- Cloud video provider: 750 PB
“We believe that data is truly going to be the driver of growth and innovation at Intel,” said Krzanich. “More of us will put out video, or high-definition images or even virtual reality. The cloud of today is built largely on people’s data.
“The cloud of the future will be built on all this machine data,” he said. “All of that data has to be pushed to the cloud. The cloud is going to take that data, apply analytics to it and push it back down to these devices. As all of these new models are applied, data is going to explode across the entire network.”
The Cloud Shift Gains Speed
The public cloud is growing faster than Intel imagined. A year ago Intel projected enterprise server CPU sales would see “mid single digit” annual growth rate. It now sees them shrinking by 5 percent on an annual basis through at least 2021, marking a “significant change in expectation,” according to Diane Bryant, the Executive VP and General Manager of Intel’s Data Center Group.
“Enterprise IT as a whole is in a significant period of transformation in a transition to cloud computing,” said Bryant. “The technology companies that serve enterprise IT are also in a period of substantial change. It’s creating a bit of a headwind for us in the near to mid term. We’re not expecting any recovery.”
Those sales are shifting to cloud service providers, which are growing at a 24 percent annual rate, and communications service providers, where annual growth is averaging 19 percent. Intel is also sharpening its focus on “adjacencies” – non-CPU data center equipment like memory and networking technology, including Silicon Photonics, 3D XPoint DIMMS, and OmniPath high-performance network fabric.
“The enterprise is weak, but it is not the swing factor in our growth,” said Bryant. “We are a data center silicon business.”
Those adjacent technologies have stronger growth than the enterprise CPU business, but also tighter profit margins. As a result, Bryant said, the data center group’s margins will contract slightly from historic levels of 45 to 50 percent to somewhere between 40 and 45 percent.
The server business may have its shifting tides, but its growth prospects remain greater than the PC market, Intel’s traditional bread and butter, which has been under siege as users shift to mobile devices like smartphones and tablets. Hence the “Data Center First” policy, which means the server business will be the first to access Intel’s newest manufacturing processes.
Bryant: ‘We Know How to Win’ in AI
Bryant focused on the opportunity for Intel in artificial intelligence (AI) computing, which represented 7 percent of Intel CPU sales in 2016. “AI is the fastest-growing segment of the data center, but it is still nascent,” said Bryant.
Intel’s AI hardware offerings include Xeon processors, the Intel Xeon Phi accelerator, and a line of FPGA accelerators using tech from its 2015 deal to buy Altera. Still to come is the highly-anticipated Lake Crest ASIC technology that Intel acquired through its Nervana acquisition, which will come to market this year.
Intel faces a stiff market challenge from NVIDIA, whose GPU technology has gained a solid foothold in high performance computing and notched early wins in AI computing, powering Facebook’s Big Sur AI servers.
Bryant says deployments using Intel chips and GPGPU acceleration account for just 7 percent of AI sales. Intel is working to support customers using NVIDIA GPUs for acceleration, she said, but aspires to lead in the AI market.
“We know how to win architectural battles,” said Bryant. “We have addressed GPGPUs in HPC before. No one else can deliver this level of integrated solutions end-to-end.”
An interesting wrinkle in Thursday’s event was Intel’s mention of development efforts in “cognitive computing,” a term closely identified with IBM’s Watson AI technology. While details are sparse, Intel clearly has aspirations in this area (see The Next Platform for more).
It Comes Down to Compute, Memory
AI is just one of the many technologies where Intel sees transformation ahead. The company dedicated parts of its Investor Day presentations to virtual reality, the Internet of Things (IoT) and autonomous cars. As these technologies grow and converge, the IT landscape will see three major changes, which were outlined by Renduchintala:
- Computing will become pervasive and ambient, in everything and everywhere.
- Compute, analytics and storage will be distributed into every fabric of the network, as data becomes aggregated locally.
- Low-latency, high-bandwidth connectivity will be the lifeblood that ties it all together.
All of these trends will boost demand for data center infrastructure and services. Intel believes that it is positioned for leadership in each of these emerging sectors.
“These are new applications, but this is all about compute and memory,” said Krzanich. “We know these areas.”