Google’s parent company Alphabet is pumping billions of dollars into its networking strategy as it works to build — and own — the artificial intelligence (AI) ecosystem, according to MTN Consulting’s Alphabet Playbook.
This report is the third in the consulting firm’s Webscale Playbook series, covering the “Super 8” webscale network operators: Alibaba, Alphabet, Amazon, Apple, Baidu, Facebook, Microsoft, and Tencent. It looks at Alphabet’s quarterly earnings and spending priorities, as well as the company’s networking strategy and disruptive impact on the network infrastructure market.
Alphabet has recently been spending more on capex than research and development as it builds data centers, submarine cables, and other network infrastructure, the report says.
For example, in February 2018, Google announced a $2.5 billion investment for 2018 alone to expand or open new data centers across the U.S. It also opened three new data centers in Switzerland and is spending $140 million on facilities in Chile. And in July 2018, Google Cloud announced plans to build its second intercontinental subsea cable network slated to go live in 2020. These activities — along with its $2.4 billion Chelsea Market deal to expand its New York campus — boosted the company’s capex spend, which peaked in the third quarter of 2018 at 17.2 percent of revenues.
Meanwhile, its research and development spend remained stable at about 15 percent of revenues for the past several quarters.
MTN Consulting says its analysis of webscale network operators shows Alphabet’s share of network and IT capex “recovered significantly” since the third quarter of 2016, to reach its peak at 17.1 percent in the last 12 months ending in the third quarter of 2018. This share is better than Amazon’s (15 percent) and Microsoft’s (12.8 percent) over the same period.
Alphabet will continue pumping money into its network infrastructure as it works to narrow the gap with the two top cloud providers Amazon and Microsoft, the report says. “Notably, Alphabet’s private subsea cable investment is seen as a move in the right direction as it could also bring benefits beyond performance and capacity,” it reads. These include reduced network vulnerability and better flexibility and cost benefits as it removes dependence on telcos operating the cable networks.
End-to-End AI Stack
The company’s aggressive networking strategy will also help it build an end-to-end AI ecosystem. This AI technology includes its edge tensor processing unit (TPU) hardware and cloud-based IoT and edge software. Google first announced its AI-focused TPUs in 2016, and launched two updated versions in the subsequent years. Additionally, Google has partnered with several chip vendors including NXP Semiconductors, Arm, Harting, Hitachi Vantara, Nexcom, and Nokia to bring its TPU kits to market.
Last October, Google announced that LG was one of the first adopters of its Edge TPUs. As part of its push to differentiate itself as the go-to cloud for AI workloads, Google has said it is the only cloud provider with integrated software and a custom hardware stack for machine learning and IoT in the cloud and at the edge.
But while it’s partnering with some chipmakers on its AI-focused hardware, the company’s progress in this space also puts it in direct competition with its existing vendor partners including Nvidia and Intel. The report says supply chain could be another challenge for Google, which doesn’t have the logistics and related services to support a large number of customers for its new Edge TPUs.