Evolving expectations
As the data landscape evolves, hyperscalers and enterprises are increasingly demanding more capacity, with 400G service becoming the new standard for rapidly interconnecting exchanges and data centers.
This growing expectation is not just about handling more traffic, but also about optimizing how crucial data infrastructure can meet the rising demands posed by AI. With the greater capacity that’s needed because of higher bandwidth resources being consumed, AI requires a strategic rethink of data center construction.
With this in mind, let’s explore the current climate for hyperscalers and enterprises – and assess how the situation is changing.
How the landscape is changing
Data centers have, to date, mostly been built in centralized urban locations – close to users to reduce latency and the mean time to cloud. This is to help provide the seamless experiences that are critical to uptake.
AI for training requires increased compute power, which means that the cost of power and proximity to renewables are now more pressing concerns for enterprises and hyperscalers.
The cost of land and space availability in dense cities are further factors. Data centers for AI are so large they don’t physically fit in places like London – and the cost of land is uneconomical.
The associated challenges have, in some cases, pushed data centers further away from the user and near unpopulated areas.
Sending mission critical data across the world to lower carbon markets such as Norway – where energy is close to 100% renewable and the climate is cooler – allows operators to reduce energy consumption and carbon footprint by 85-95%.
Addressing the double-edged sword
While placing data centers far away has its benefits, it also creates a requirement to securely and reliably transport data further distances. Subsea connectivity is the obvious answer, but it can be concerning for hyperscalers and enterprises.
Not only is it challenging to transmit mission critical data across large distances in this way, but it also relies on having a large number of routes and landing points to handle the growing volume of traffic. It means that gaining the support of a trusted partner who can offer a wide range of diverse routing options is critical.
Data center emissions
The International Data Corporation (IDC) projects datacenter energy consumption to grow at a 16.0% CAGR from 382TWh in 2022 to 803TWh in 20271. This is another reason why low carbon markets are being increasingly explored to help reduce impact.
Source: Climatiq Analysis, The Shift Project, Measuring greenhouse gas emissions in data centres: the environmental impact of cloud computing, April 2022
Carbon intensity of data centers
When it comes to carbon intensity, hyperscalers and co-located data centers are significantly more efficient than internal data centers – meaning a move away from them would help contribute toward better long-term sustainability.
Source: Md Abu Bakar Siddik, Arman Shehabi, Landon Marston, The environmental footprint of data centers in the United States, Jan 2021
Carbon intensity of data centers
When it comes to carbon intensity, hyperscalers and co-located data centers are significantly more efficient than internal data centers – meaning a move away from them would help contribute toward better long-term sustainability.
Source: Siddik & Sehab 2021
1 IDC, Datacenter Energy and Carbon Emission Projections and the Drive to Net Zero, June 2023
