Google Expands Cloud Data Center Plans, Asserts Hardware, Connectivity Leadership

Company touts its infrastructure might as it fights for enterprise cloud market share

Yevgeniy Sverdlik

March 9, 2017

5 Min Read
Urs Hölzle, Senior Vice President for Technical Infrastructure at Google, speaks during the Google I/O 2014 conference in San Francisco
Urs Hölzle, Senior Vice President for Technical Infrastructure at Google, speaks during the Google I/O 2014 conference in San FranciscoStephen Lam/Getty Images

Google has added three locations to the list of cloud data center construction projects it is doing to expand geographic reach of the physical infrastructure that supports its enterprise cloud services, investing billions to catch up with its biggest rivals in the space -- Amazon, and Microsoft.

Urs Hölzle, the company’s senior VP for technical infrastructure, announced the plans from stage Thursday at the Google Cloud Next event in San Francisco. He also underlined hardware innovation inside Google’s cloud data centers and its global network backbone and announced new cost controls for cloud users, while his colleagues unveiled new security, developer, and collaboration tools.

Cloud data center locations, connectivity, hardware, cost, security, and toolset are all levers cloud providers pull as they compete for enterprise cloud market share – a race in which Google Cloud Platform remains far behind market leader Amazon Web Services as well as Amazon’s distant-second competitor, Microsoft Azure.

AWS had more than 40 percent market share in public infrastructure and platform cloud services at the end of last year, Synergy Research Group estimated. Microsoft, Google, and IBM collectively had 23 percent. Synergy doesn’t break out the latter three companies’ individual market share, but another market analyst firm, Structure Research, estimated that at least in the infrastructure-as-a-service space, Azure commanded close to 11 percent market share in 2015, while Google’s share was 2.5 percent, compared to AWS’s nearly 71 percent.

DCK analysis: Can Google Lure More Enterprises Inside Its Data Centers?

But Google’s cloud business is growing. The company paraded numerous big new customers at the conference, including Verizon, Colgate-Palmolive, HSBC, eBay, and Evernote, among others. “Customers of GCP connect to a billion individual users every single day,” Hölzle said.

Expanding the Google Data Center Empire

The three new cloud data center locations Google is planning to launch are in California, Canada, and The Netherlands, Hölzle said. Previously announced regions that are in the works are Northern Virginia, São Paulo, London, Finland, Frankfurt, Mumbai, Singapore, and Sydney.

All these locations will come online this or next year, he said, bringing Google’s total to 17 availability regions and 50 availability zones. Each zone is essentially a separate data center, with its own dedicated power and cooling infrastructure. The company plans to have a minimum of three zones in each region, although some regions are initially launched with two zones.

Like its big rivals, Google has been spending billions of dollars each year on building out its global network, which includes data centers, hardware, and telecommunications infrastructure, including terrestrial and submarine fiber cables. Its trailing three-year capital investment totals about $26.4 billion, Hölzle said. Capex figures Google discloses usually include costs other than infrastructure, but infrastructure is responsible for the bulk of the expenses.

The company reported $10.9 billion in capex for 2016 – up 10 percent from the year before, an increase that was in line with its recently expanded focus on cloud infrastructure investment.

Chipping Away at Market Share

One of the layers of the stack where Google is asserting leadership is data center hardware. In addition to designing its own servers and network switches, Google has been deeply involved in the design of CPUs that power its cloud services by collaborating with Intel, which supplies chips for virtually all computing capacity cloud providers have today.

Last month, Google announced that it was the first company to upgrade its servers with Intel’s next-generation Xeon processors, codenamed Skylake. At the event Thursday, Raejeanne Skillern, who leads Intel’s Cloud Service Provider Group, confirmed that Google was first, and that Intel would not be putting the product on the market for some time.

While it’s common practice for Intel to customize products for all hyper-scale cloud companies – it’s been tweaking its chips for Google since 2003 – Google was involved in the development of Skylake from the very beginning. “We take the Google feedback every step of the way,” Skillern said.

Hölzle also shared more details about a custom security chip that’s on every server motherboard in Google data centers, which was only briefly mentioned in a recently published whitepaper on the company’s cloud security practices. The chip is called Titan, and it’s so tiny that Hölzle was able to have it attached to an earring pin he was wearing on stage. Titan helps Google protect hardware at the BIOS layer. It helps authenticate hardware and services running on the servers.

Wiring the Planet

Another big capital investment area for Google is global data center connectivity. It was the first non-telco technology company to invest in a submarine cable construction project nine years ago. The cable, called Unity, crosses the Pacific Ocean to link landing stations in Chikura, Japan, and Redondo Beach, California. It came online in 2010.

Since then, Google has invested in five more submarine cable projects. In recent years, other hyper-scale data center operators, including Microsoft, Amazon, and Facebook, have also become major investors in transcontinental connectivity.

Read more: Here are the Submarine Cables Funded by Cloud Giants

Sharpening the Machine Learning Angle

Google has been making good progress on differentiating itself with cloud services around Artificial Intelligence and machine learning (the most widely used type of AI), Sid Nag, a research director at Gartner, said. That has been the company’s core message since around one year ago, when it started significantly ramping up its enterprise cloud efforts.

A key challenge going forward will be simplifying the narrative around these tools and creating some “pedestrian use cases” to help customers easily integrate Google’s machine learning capabilities into their applications, Nag said.

Go-to-Market Improvements

Another important development this week was Google’s announcement of a significant expansion of its channel and technology partner programs. The company hasn’t had a managed services capability to help clients get on-boarded onto its cloud platform, and partnerships such as the one it announced with Rackspace, will go a long way in improving that.

The managed service provider has been supporting AWS and Azure, saying these managed cloud services have been its fastest-growing business. Now, it has added the third massive cloud platform into the mix. Google’s “partnership with Rackspace is going to be pretty significant,” Nag said.

Read more about:

Google Alphabet
Subscribe to the Data Center Knowledge Newsletter
Get analysis and expert insight on the latest in data center business and technology delivered to your inbox daily.

You May Also Like