Cloud growth and development is entering the most dynamic period we have ever seen. Hyperscale cloud companies are spending USD 100 billion per year on CAPX, growing over 20 percent per year. This is driving major productivity improvements in infrastructure and AI-based applications. We are likely only 20 percent of the way through cloud adoption.
Data is the key commodity driving AI, which almost all new cloud applications are taking advantage of. However, AI will require a new infrastructure design and retrofitting of legacy data centers with much more network and compute density.
Bottlenecks have shifted from last mile to middle mile and specifically within data center interconnectivity. New builds are moving outside of city centers to take advantage of low-cost power and land. This infrastructure is driving down costs and enabling new services such as Software as a Service, TikTok, Disney+, and now an explosion of work, entertainment, and learning at home on a global basis.
Gigawatt-sized data centers are being built that need to connect back to existing data center campuses to ensure low latency to data oceans. 17,000 pairs of fiber are now installed by a single large customer with all of the strands lit on day one.
Enterprise customers are looking for scalable and flexible infrastructure to handle huge levels of complexity that they never expected. Green energy is now becoming a must. Dark fiber, green power, and land are all in major demand and remain bottlenecks in a very high-demand environment.
At PTC’20, I moderated a four-person panel discussion about cloud infrastructure highlights. Michael Foust from Chindata Group, Craig Scroggie from NEXTDC, Chris Sharp from Digital Realty, and Toby Williams from euNetworks. The first three panelists focused on data centers while Toby provided his perspective on wholesale fiber bandwidth primarily connecting data centers.