SANTA CLARA, Calif. -- Intel's claim to have reinvented itself as a "data-centric" company is baffling at first. Hasn't Intel, and every computer vendor, always been "data-centric?" Computing has been data centric since Herman Hollerith first used punch cards to record data well more than a century ago, in the 1890 US Census.

Yes, computing has always been about the data, but today's emerging applications, in areas such as the Internet of Things and AI, require data in quantities and at speeds that paralyze previous technologies. Today's computing requires new hardware, storage and networking -- fundamentally different from anything available previously, according to Intel.

Intel's transition is also driven by a change in its own business. The client computing business is becoming less important, and the company needs to change to survive. So you can view Intel's "data-centric" business as "everything that's not the PC and mobile."

"We are hungry to get after this market in 2018, 2019 and beyond," said Navin Shenoy, executive vice president and general manager of the Data Center Group at Intel, kicking off a day of briefings for press and analysts on Wednesday. Five years ago, roughly a third of Intel's revenue was data-centric; now it's half, Shenoy said.

Figure 1: Intel's Navin Shenoy

Intel's data-centric business grew 26% year-over-year to $8 billion in the second quarter ending June 30. The PC-centric business grew too, but by a more modest 6%, to $8.7 billion. While the PC business still accounts for the majority of Intel's business, it is only just the biggest revenue generator, with 51% of sales versus 49% for the data-centric business. Overall, Intel grew revenue 15% in the second quarter, to $17 billion. (See Intel reports $17B in Q2 revenue, up 15% YoY.)

But not everything is rosy for Intel.

Intel is staking a big part of its future on AI, and it's late to market there, fighting incumbent Nvidia Corp. (Nasdaq: NVDA) and several emerging companies. (See Intel Reinventing Xeon for AI – but Is It Too Late?)

And it also faces a challenge from Advanced Micro Devices Inc. (NYSE: AMD). "If AMD can get into Google and Amazon, Intel will be in trouble," Gartner analyst Martin Reynolds tells Light Reading. Hypercloud providers would like to see more choice, to help drive better terms and pricing. "The big guys would like a lever," Reynolds says. He sees the probability of that happening at about 30%.

Autonomous cars will run on data
Intel defines the data-centric business as products and technology powering the cloud, data center, enterprise, the edge, Internet of Things, 5G and networks. It includes AI, Field-Programmable Gate Arrays (FPGAs), data center memory and emerging silicon photonics networking -- technologies to "process, analyze, store and move data to drive business value," Shenoy said.

The data-centric business specifically excludes Intel's traditional strength in desktop and laptop processors, although those businesses remain important to the company.

The transportation industry, which is ripe for disruption, is a good example of the data-centric transformation, Shenoy said. The autonomous car, in particular, is "an end-to-end-compute process which exercises all the assets of Intel…. Autonomous vehicles will run on data, just like today's vehicles run on gasoline."

Intel got into the autonomous car industry in March 2017 with the $15.3 billion acquisition of Mobileye, based in Israel, which had been developing self-driving technologies for nearly 20 years. (See Intel, Mobileye $15.3B deal has cloud under the hood.)

An autonomous car requires high-definition maps, both located in the cloud and gathered in real time from cameras located on the vehicle, Shenoy said. Each car generates 4TB of data per hour. The car's own data is sent to the cloud, and combined with data from other vehicles to create models, requiring increased need for cloud computing resources.

"And then once the models are deployed they have to blend intelligence in the cloud with intelligence and vision streams in the car to enable continuously updating realtime high definition maps," Shenoy said.

The car will require ten times the intelligence of today's cars, and compute and network intelligence on the network edge to handle delivering data and intelligence in an efficient way, Shenoy said.

"As the car is driving down the highway, it will need to be continuously updated," Shenoy said. The car itself has only 20–40 kilometers of data on board.

And that's just one industry. Altogether, the change to data-centric computing represents a $200 billion total addressable market by 2022 -- add client and mobile, and the market is north of $300 billion, Shenoy said.

Next page: The enterprise slump

Intel has 20% of the data centric market, with plenty of opportunity for growth, Shenoy said.

Into that market, Intel plans to sell the next generation of its venerable Xeon processor, as well as emerging technologies from Intel such as silicon photonics, smart NICs, custom ASICs, FPGAs and Optane permanent memory.

Three big drivers
The cloud, along with networking and AI, are the first big drivers of data-centric networking.

Intel's cloud business is growing at a rate of 41% year-over-year this year, Shenoy said. Cloud providers demand custom CPUs from Intel. Five years ago, 18% of CPU volume to cloud service providers was custom, now half is custom.

Communication services providers are the second driver for data centric computing, including 5G and edge computing. The cloudification of service provider networks will be a $24 billion market by 2022, and Intel has about 20% of that market today. Communications service providers are moving from proprietary equipment to standard, high-volume servers, Shenoy noted. (See Telcos Must Go Cloud-Native to Compete .)

AI is another data-centric trend, with a $2.5 billion silicon market today, growing at a double-digit rate to $10 billion in 2022. AI is Intel's single biggest investment; Intel is building purpose-built AI products and also building AI into all its products.

Data-centric computing requires advances in storage technology; businesses need somewhere to put all that data where it can be accessed quickly. To that end, Intel is launching a new line of Optane persistent data center memory, providing storage for applications such as Spark SQL and Apache Cassandra. As the name implies, persistent storage keeps data intact when power is lost. High availability systems can go from minutes in reboot and restart times to seconds -- three nines to five nines of availability, Shenoy said. The first production units of Optane persistent memory shipped from Intel's factories Tuesday.

Google Cloud Platform is partnering with Intel to deliver Optane persistent memory to Google's cloud customers. Other support comes from Alibaba, Cisco, Fujitsu, Dell EMC, Hewlett Packard Enterprise Microsoft, Oracle, Red Hat and more.

As part of its transition, Intel is recovering from a slump in its core enterprise market. Sales lagged in 2014–17, driven by a perception then that all workloads would be moving to the public cloud, said Raj Hazra, Intel corporate vice president, data center group. But now, Intel -- and enterprises -- see a strong future for hybrid and private cloud, with AI driving on-premises infrastructure growth.

During the slump, Intel worked with cloud and enterprise leaders, including Microsoft Azure, Google Cloud Platform, Red Hat and VMware, on hybrid and private cloud architectures, Hazra said. Intel has been expanding its analytics investment and growing its AI business, in partnership with vendors including Cloudera, Oracle, SAP and SAS.

Currently, Intel's data center strategy is driven by enterprise demand for private cloud, AI and "repatriation" -- enterprises moving workloads from public cloud back to private cloud, even as they continue to grow their public cloud deployments, Hazra said.

Private cloud is adopted by 12% of enterprises today, compared with 6% five years ago. Some 80% of enterprises are repatriating workloads to private clouds. And according to Hazra, AI and analytics on-prem CPU deployment is seeing a 2x growth rate from 2017–2021 compared with 2014–2016.

"The period of confusion is largely over," Hazra said. And enterprise demand for servers has picked up.

Related posts:

— Mitch Wagner Executive Editor, Light Reading