A brand new report from Moody’s Rankings highlights the dramatic enhance in information heart demand, primarily fueled by developments in synthetic intelligence (AI) and ongoing digital transformation.
The report, titled ‘Data Centers – Synthetic Intelligence: Tech Giants’ Speedy Buildout of Data Centers to Meet AI Demand is Not With out Threat,’ supplies insights into the present state and future projections of the worldwide information heart trade.
Among the many report’s key findings:
-
International information heart capability is anticipated to double over the subsequent 5 years
-
Knowledge heart electrical energy consumption is forecast to develop by a median of 23% yearly between 2023 and 2028
-
AI-specific information heart power utilization is projected to develop by a median of 43% yearly over the identical interval
-
Hyperscalers estimated to extend annual IT spending by $48 billion in 2024
The dramatic spike in information heart progress was not surprising to Moody’s. “We aren’t in any respect stunned by the impression of AI on information heart capability, as GPUs processing AI workloads require a lot increased energy consumption in comparison with the standard processors,” Raj Joshi, senior vp for Moody’s Rankings, instructed Data Center Information.
“The dimensions of latest AI fashions is rising quickly and adoption charges of AI are very excessive, so we have been stunned by the pace of AI adoption and the tempo of expertise innovation, that are driving the surge in information heart infrastructure spending.”
The place All of the Data Center Energy is Going
The Moody’s report (registration required) notes that demand for information heart capability is surging as a result of computational energy wanted for AI developments and ongoing digital transformation.
Hyperscalers together with Amazon, Google, Microsoft, and Meta are quickly constructing and leasing new information heart capability to fulfill anticipated future demand, specializing in each house and energy.
With AI there are two main operations: ‘coaching’, the place fashions are constructed and expanded over time, and ‘inference’, the place present fashions are used to derive a response.
Based on Moody’s, significantly much less computing assets are required for inferencing than coaching. Nonetheless the report goes on to notice that “inferencing is rising rapidly and the amount of inferencing requests to information facilities will develop considerably as utilization of AI-powered purposes expands with rising adoption.”
Moody’s mentioned it expects that inferencing will develop to signify the overwhelming majority of AI workloads over the subsequent 5 years.
Cloud Repatriation is Not a Main Concern
Some studies, together with latest perception from CoreSite, declare that cloud repatriation is a rising concern. Cloud repatriation nevertheless isn’t worrying Moody’s a lot, if in any respect.
“Repatriation of workloads from the cloud just isn’t new however migration of workloads to the cloud far exceeds the repatriation volumes,” Joshi mentioned. “That is evidenced by progress charges of hyperscalers, subdued progress charges of IT {hardware} infrastructure distributors that primarily serve the on-prem markets, and numerous third-party studies.”
Cloud Hyperscalers will Prioritize AI Investments
Moody’s isn’t the one group forecasting robust information heart and cloud progress. A latest report from BlackRock additionally famous robust expectations for information heart progress within the subsequent few years.
Dell’Oro Group senior director Baron Fung has a nuanced view of how information heart progress will persist.
“I agree that demand for conventional cloud companies will proceed to drive strong demand,” Fung instructed Data Center Information.
Fung added that there was a little bit of a correction cycle final yr through which enterprise prospects optimized their cloud utilization. He famous that some cloud service suppliers, resembling AWS, are actually seeing extra stability. Nonetheless, wanting forward, he warned that the market ought to count on decrease progress in cloud revenues for conventional companies than was seen a number of years in the past.
General, Fung expects that the expansion price can be within the strong double-digit annual progress vary.
“Trying forward, we should always proceed to see massive enterprise prospects transfer selective workloads to the general public cloud, and SMBs will rely extra on the general public cloud,” Fung mentioned. “Nonetheless, the market is approaching an equilibrium level through which some quantity of workloads will stay in personal information facilities on-premises, somewhat than the cloud.”
Fung additionally expects that cloud service suppliers will proceed to prioritize AI investments over conventional IT investments by means of this yr.
“It’s quite a lot of funding, and there will not be a return on funding for a while, however nobody needs to be left behind,” Fung mentioned. “First adopters have a tendency to keep up a powerful benefit, like Amazon did with the cloud round 10 years in the past.”
