Intel’s Diane Bryant Blazing Trail From Data Centers to Next Era of IT
In a world of apps, big data, and cloud connectivity, it can be hard to remember that there’s more to tech innovation than just software. Then you meet someone like Diane Bryant, and your worldview shifts.
Bryant is senior vice president and general manager of the datacenter and connected systems group at Intel. You know, the $50 billion chipmaker that holds the hardware keys to not just PCs, but the data centers and servers that power our 21st century digital lifestyle. Bryant runs Intel’s fastest-growing business unit—responsible for 22 percent of the company’s revenues and 33 percent of profits last year—and she predicts it will double from a $10 billion business (in 2011) to $20 billion by 2016.
Santa Clara, CA-based Intel (NASDAQ: INTC) has been widely criticized for being slow to move into chips for smartphones and other mobile devices. That market currently makes up less than 5 percent of the company’s revenue. Its cash cow—something like three-quarters of its sales—lies in PCs, laptops, and (increasingly) tablets. Still, the global market for PCs is shrinking.
Meanwhile, Intel is also going through a major leadership transition. Former CEO Paul Otellini, who ran the company since 2005, stepped down in May, and former chief operating officer Brian Krzanich has taken his place. And late last month, CTO Justin Rattner, who led Intel Labs, stepped down because of an age requirement for Intel officers (he turned 65 and is slated to return in a different role). Krzanich spoke recently about ramping up the company’s efforts in mobile devices; Intel also made its first acquisition on his watch, of a satellite navigation chip business. (The company is set to announce quarterly financials this week.)
But data centers are where a lot of the action and growth are these days. Internet giants like Google, Facebook, Amazon, and Microsoft depend on Intel’s technology—often customized processors and systems, jointly developed—for their massive infrastructure needs. When you do a search, post a social-media update, or shop online, your request is processed, ultimately, by a piece of hardware sitting in a server. And as more machines and people get networked together—whether you call it the Internet of things, connected devices, or M2M (machine to machine)—the number of server requests is going up and to the right.
It all adds up to a huge emerging opportunity for Intel—and for Bryant’s group in particular.
Last month, I met up with Bryant in the Boston area, ahead of a private roundtable event on big data in finance. I wanted to get a sense of the strategic thinking that goes into Intel’s decisions about where to invest, how to compete more effectively in IT, and how the company deals with everything from energy efficiency and security to large-scale manufacturing issues.
Bryant has a pretty interesting perspective. She went to work for Intel as an engineer straight out of college in 1985. She walked me through the tech trends of each subsequent decade:
The ‘90s, she says, were computer-centric; they were about technology deployment and automation—making systems go from manual to computerized (think PCs). The 2000s were network-centric; they were about connections and the rise of the Internet and cloud. The present decade is what she calls “human-centric” and “service oriented.” Now tech companies are “using IT to deliver services directly,” she says, through advances in cloud computing, big data, and high-performance computing.
What’s more, the overall role of information technology has changed radically, she says. Bryant would know; she oversaw Intel’s IT organization as chief information officer for about four years before moving to her current post in early 2012. “IT used to support the business. Now IT is the business,” she says.
A look at the server-chip market trends is pretty instructive. Intel’s sales to big Web companies—what Bryant calls “public cloud service providers” like Google, Facebook, and Amazon—are growing at an annual clip of 25 percent, she says. That’s much stronger growth than in Intel’s more traditional market segments, such as enterprise IT (Oracle, HP, IBM, Dell), which is at 8 percent, and telecom service providers, which is at 12 to 15 percent. (High-performance computing is another growth area, mostly for government customers, at about 20 percent.)
The trend suggests where IT resources as a whole are flowing—to more Web and consumer-facing businesses, at the expense of more traditional IT providers. “Everyone has to shift their business model or suffer,” Bryant says.
So what are the biggest issues Intel is dealing with now?
One is energy efficiency. A lot has been made of the power requirements of data centers—something like 25 percent of the total operating cost of a data center is electricity, and much of that goes into cooling down hardware. Bryant asserts that Intel’s transistors “run at better energy efficiency than any other.” No doubt its competitors would differ, but at least in server chips, Intel says it has gotten the power requirements of its Atom line (lower performance) down below 6 watts, and its Xeon line (higher performance) down to about 13 watts.
Intel has also been working with Facebook to redefine and standardize data-center equipment at the server-rack level. As part of Facebook’s Open Compute Project, the companies have developed a reference design that tries to eliminate redundant pieces of hardware in a given server rack by aggregating different pieces of processing, storage, and networking in a more efficient way. (That won’t make vendors like Cisco or EMC happy, but if done right it could lead to more efficient data centers.)
Another big issue these days is cybersecurity. Intel’s $7.7 billion acquisition of McAfee in 2010 has led the company to hardwire more anti-virus features and malware blockers into its newer PC and server chips. The technology can do things like shut off applications when a server is compromised, Bryant says. Intel also has been working on better encryption tools to improve data security in the cloud.
In a similar vein, the company has embedded what it calls “trusted execution technology” into its Xeon server processors. This is an effort to create pools of trusted servers within virtualized data centers, so that IT administrators can keep track of which servers are secure and which may be compromised. Sounds obvious, but I think the key lies in how the system uses measurements to verify the condition of both hardware and software in real time.
Meanwhile, for all its progress, Intel is still in the chipmaking business—which means it deals in the global-scale advantages and headaches of manufacturing. Last I checked, the company had fabrication plants as far-flung as Arizona, Oregon, Israel, and China (and even more test and assembly sites). I asked Bryant how the company tries to press its advantage in fabs.
As she puts it, Intel and Samsung are basically the only players left that have their own chip factories. Qualcomm, Nvidia, and Toshiba outsource their manufacturing, and Intel’s archrival Advanced Micro Devices (AMD) divested its manufacturing business in 2009, which became GlobalFoundries. Intel has been a clear leader in cramming more transistors and processing power onto smaller and smaller surfaces; this year it will be among the first to move to the 14-nanometer process (which refers to the tiny scale at which chip substrates are etched).
Bryant credits Intel’s multibillion-dollar investment in R&D for that. I took that to mean both the billions it takes to build each fab, and paying for research into the advanced physics and processes of manufacturing chips at such a scale. (I also take it that working with an independent fab as a chip designer would have major drawbacks, though it’s much cheaper.)
Yet the competitive environment in servers—and elsewhere—remains treacherous. “We take all competitors seriously,” Bryant says.
One class of competitors stands out. All told, she says, there are about 15 hardware companies that are developing ARM-based processors for servers. That’s the same ARM Holdings whose chip architecture, known for its low power consumption, has come to dominate the mobile-device chip market. And one of the companies leading the way in ARM-based server chips is—you guessed it—Intel nemesis AMD.
Bryant shot down a series of what she called “myths” about ARM-based server chips—that they are more energy efficient, easily compatible with existing data centers, and so forth. She thinks all the different kinds of processing units (cores) put out by the 15-odd vendors will be a big disadvantage. “That variation is what’s going to kill them,” she says. In the data-center market, “variation is your enemy. It costs [manufacturers] money.”
In any case, none of the competition seems to dissuade Bryant from her main conclusion that Intel will continue to ride a huge wave of Internet-connected devices and cloud services into the next era of tech. The main question in my mind—and this is outside Bryant’s domain for now—is whether Intel’s chips will find their way into enough end-user devices as well.
“It’s a very young industry,” she says. “It’s so dynamic.”