Amazon CTO Werner Vogels: 'Infrastructure is not a differentiator'

NEW YORK -- It just hung there, in the air.

An uncomfortable truth, hovering before the faces of some 3,000 technologists sitting in the shadows to hear the opening keynote here at the Amazon Web Services Summit 2013.

Amazon chief technology officer Werner Vogels was standing on the expansive stage, clad turtleneck to toes in black, delivering his speech as scheduled. But in a sea of predictable, self-congratulatory praise—the kind of rah-rah enthusiasm endemic in the technology industry—he dished out a surprisingly cutting comment.

"You have to stop wasting time and effort for things that do not matter for your customer. One of those things is your infrastructure. It is not a differentiator."

A few people shifted in their seats. Some exhaled nervously. A number of them blinked, the only sign of comprehension on an otherwise blank visage.

Not 20 minutes into his 1.5-hour keynote presentation, the most prominent speaker on the schedule told an audience of developers, project managers and technology executives that what they did for a living does not matter. Because Amazon, undeniably, could do it better at scale.

Vogels spent some time making his case. His company has tackled the cloud computing market with the kind of zeal that Walmart has approached brick-and-mortar retail. Its ardent pursuit of efficiencies has produced 31 price reductions since 2006, he said.

"This sort of flywheel helps you with economies of scale and drives our costs down," he said. "We've made the decision to pass those benefits back to you [in reduced costs]."

The numbers are staggering. Amazon S3 has surpassed two trillion objects and routinely runs more than 1.1 million peak requests per second. The company's Elastic MapReduce has launched 5.5 million clusters since its launch in May 2010.

And the company continues to keep pace with market changes: Vogel took his speaking opportunity to announce new Local Secondary Indexes for Amazon's DynamoDB product and transparent data encryption and native network encryption for Amazon RDS for Oracle, available today worldwide.

"Why are they doing this?" Vogels asked the audience rhetorically. "[Because] the database has always—always—been the bottleneck in this game." Why not instead request a certain amount of performance from the database, let it figure out that amount for itself and then have it delivered with consistently low latency?

"If we are able to drive the cost of this compute and storage down to a point where you no longer have to think about it, tremendous products are going to be built," Vogels said. "You will no longer be constrained by your infrastructure." 

"No longer is the database the bottleneck. It shifts into your applications."

It is undeniable that everything Vogels said is in Amazon's own interest. More customers lead to more revenues and more efficiencies at a greater scale. But he's not wrong: it may be in vogue to be a "technology company," whether for the healthcare or publishing or financial services or hospitality industries, but it is an awfully expensive proposition. And, if Amazon continues its cloudy mission unabated, there will be a point where it will be economically irrational to manage one's own infrastructure—nevermind the fact that the company will probably be better at it than you.

"To increase innovation, reduce the cost of failure," Vogels said, citing MIT Media Lab director Joi Ito. "The cost of failure here is close to zero."

The industry angle

To bolster his point, Vogels ceded the stage to several customers, who offered the audience their own use case, through the lens of a specific industry.

Russell Towell, a senior solutions specialist for Bristol-Myers Squibb, explained how his company used Amazon's services to build a research cloud. "When you can optimize and give these scientists the ability to run thousands and thousands of clinical trials, they can optimize design," he said, thereby reducing the number of human subjects used in a study, reducing its length and frequency of data collection, and ultimately cost to conduct.

Vogels cited the insurance industry—"a price-sensitive, highly competitive, data-driven" industry with "huge capacity requirements that are only needed a half-day a week"—as an example of a company that is radically changing the way it works with a cloud-based approach that emphasizes data-based modeling and removes the need for on-site visits.

Scott Mullins, a managing director for Nasdaq OMX, explained how his company used the cloud to manage one-tenth of all trading transactions worldwide. Joining him was Holly Hasty, a senior vice president for U.S. bank FirstSouthwest, who explained how her company used the cloud to manage legacy data as it switched back-office system vendors.

Joe Salvo, a manager at GE Global Research, explained how his company built its own platform, called "Crowd-driven Ecosystem for Evolutionary Design," to speed up the design-manufacturing process for its global operations as complexity for the products it makes—locomotives, wind turbines, power turbines, health equipment—increases.

Mortar Data chief executive K Young explained how his small startup was able to take on less financing ($1.8 million, versus an estimated $7.1 million) and scale much more quickly by building in the cloud.

Vogels also cited the biotechnology and hospitality industries. In the former, consumer goods conglomerate Unilever is (using the Eagle Genomics platform) looking for microbes in the human armpit, triggered by particular genes, so that it can build better deodorant, he said. In the latter, the European luxury hotel chain Kempinski is focusing on core competencies.

"They realized that every [dollar] that they can spend on customers, and not do heavy lifting for IT, is important," Vogels said. "They realized that customer services is going to win over the customer, not IT. So they're moving everything into AWS."

The examples continued: Vogels said that Shell, the energy company, wants to deploy 10,000 experimental sensors, each which generates petabytes of data, around the world. "Their data needs are becoming gigantic -- something you'd never be able to manage on-premesis." GE and Mount Sinai Hospital partnered to place RFID tags on every patient and item to optimize patient flow and asset management, ultimately treating 10,000 more patients and saving $200 million.

His point? In the age of cloud computing, whatever the industry, building and managing the infrastructure to support the above applications is either a distraction at best or prohibitively expensive at worst.

Infrastructure is not a differentiator. (Unless you're Amazon, that is.)

"This whole world is radically different," Vogels said. "Cloud computing has changed the data business."