In Blog

Cloud growth and development is in its early stages, and it is perceived that AI will accelerate this process. There is a huge runway from cloud and the transition of IT to AI.

Cloud services revenue amounts to USD 100 billion and is expected to triple within six years, inside a USD 3 trillion global IT spend. AI will be a big part of this growth. The industry will need to invest USD 25 billion in the data center infrastructure to support AI alone, a big increase from the historical data center spend in the USD 7 billion range. Data is the key commodity to drive AI. Consumer-created data will be 10 times what enterprise is and machine-created data will be 50 times what enterprise data is. This much data will create a gravity effect, where compute has to go to where the data is stored (data lakes) because it will be hard to take data to the compute. With data centers, location is critical. In addition, logical locations for the Internet will have to shift, which will require new data networking protocols to support this.

The role of the data center has changed dramatically in the last 15 years. Historically it was enterprise-driven and the data centers were the destination. Now the cloud is the destination.  Hyperscale cloud providers are now 70 percent of the incremental growth in the industry. These companies spent USD 90 billion in infrastructure in 2018, up from USD 20 billion in 2012, and require more global infrastructure with better coverage in countries of smaller cities.

These data centers are becoming interconnected globally and enterprises are using the hyperscale providers much more. Cloud network traffic has been increasingly occurring within and between data centers. Networking is now about getting critical cloud resources. They are adopting the same virtualized technologies that the hyperscale Internet companies use, Software-defined Networking (SDN), which have been a game changer for networking, reducing costs and complexity, and improving quality.

AI is driving new cutting-edge computing infrastructure – high performance computing clusters with customized installations and maintenance. These are new workloads that require high density computing clusters, new types of liquid cooling, and more energy- dense data centers. Very few people have the technical ability to deploy AI cloud capabilities.

At PTC’19, I had the pleasure of moderating a panel discussion on this very topic. Take a look as to what Raul Martynek, chief executive officer of DataBank and Jonathan Schildkraut, EVP & chief strategy officer of CyrusOne had to share about data infrastructure needs of the 21st century.