E D I T O R ' S Q U E S T I O N
ANDREW FRAY , HEAD OF EMEA DATA CENTERS , CUSHMAN & WAKEFIELD
As an elder statesman of the data centre industry , I am frequently asked whether Edge Computing is an attractive investment , since ‘ the Edge ’ is where carrier neutral data centres started nearly 35 years ago ; when London Docklands for Telehouse and then Frankfurt Gallus for Ancotel became key interconnection points for deregulated voice telecommunications .
It ’ s fair to say that the main drivers for investors to deploy capital and infrastructure into markets outside the traditional FLAP-D have been the major cloud service providers , seeking to establish new availability zones to cover enterprises , consumers and ultimately GDP across EMEA . Such regions are influenced by data protection ( especially GDPR ) in Europe , as well as the availability of suitable land , cheap and renewable power , and the complex ‘ signals ’ which the hyperscalers use to determine where to deploy next .
Over the last few years the demand from hyperscale has absorbed in excess of 85 % of the available , under construction , or planned data centres in EMEA .
But times are changing , as cloud computing , while growing by low-tomiddle double-digits per annum , is being supplemented by the ‘ megascale ’ requirements for Artificial Intelligence . At the current time , huge sites with massive power capability are being sought across the region in a form of a modern ‘ Klondike Gold Rush ’, with per site requirements 5 – 10x what they were 18 months ago .
So where does this leave Edge Computing ? The current power land grab for learning AI will be followed by the need to provide data centres for inference AI . These centres are likely to be highly-powered , smaller and , for latency reasons , placed in urban areas adjacent to end users and workloads .
On one one hand , the future looks very bright as Edge includes the data interconnection , peering points , towers and landing stations that it ever did , and with global connectivity strengthening , this means that the Edge will be found in all countries . As I used to prophesise in the distant past ‘ every major city needs a carrier neutral data centre ’ and I still stand by that . Both B2B and B2C applications need delivery at the Edge to run and thrive .
On the flip side , the case for locally generated IT workloads being processed locally around relatively unattractive ( from a data gravity perspective ) metro areas is , in my view , still unconvincing . Whilst the disaggregation of data seems inevitable , in the same way that the desktop PC augmented the mainframe computer , I remain uneasy about the data centres currently in scope for this ‘ enterprise
“
THE CURRENT POWER LAND GRAB FOR LEARNING AI WILL BE FOLLOWED BY THE NEED TO PROVIDE DATA CENTRES FOR INFERENCE AI . THESE CENTRES ARE LIKELY TO BE HIGHLY-POWERED , SMALLER AND , FOR LATENCY REASONS , PLACED IN URBAN AREAS ADJACENT TO END USERS AND WORKLOADS .
Edge ’, since they are frequently older , legacy , high cost , low efficiency and lower powered sites . The success of these data centres is influenced by many workloads still being migrated to standard public cloud , such as AWS , Google Cloud , Microsoft Azure , Oracle Cloud , Salesforce and others .
I therefore suspect that we will see Edge data centres deployed as part of ‘ everything , everywhere all at once ’, but not in the way they are currently deployed . To mis-quote Star Trek : ‘ There ’ ll be data centres , Jim , but not as we know it ’. �
36 www . intelligentdatacentres . com