Knowledge facilities elevated capital expenditures (capex) by almost 50% within the second quarter of 2024, pushed by the urge for food for AI-ready {hardware}, in line with a brand new report from Dell’Oro Group.
Perception from the California-based analysis agency signifies that information heart capex grew by 46% within the second quarter of 2024, in comparison with the identical interval a 12 months earlier.
The spending spree is predicted to proceed for the rest of the 12 months, albeit at a barely slower tempo. Dell’Oro forecasts a sturdy 35% progress in worldwide information heart capex total for 2024.
The group mentioned the expansion is basically fueled by the growing demand for GPU-accelerated servers designed for AI purposes.
“Accelerated servers are servers with GPUs and different customized accelerators, such because the Google TPU,” Baron Fung, senior analysis director at Dell’Oro Group, instructed Data Center Information.
The Dell’Oro numbers are the newest in a collection of double-digit information heart expenditure progress predictions. Gartner forecasts that total spending on information heart methods will develop by 24% in 2024. Like Dell’Oro, progress is partly attributed to the necessity to energy generative AI methods and providers.
Omdia forecast on the finish of 2023 that the worldwide server market could be price $196 billion by 2027, whereas, a report from Elementary Enterprise Insights & Consulting launched earlier this month estimated the marketplace for AI servers to be price $183 billion by 2032.
AI Acceleration: The Engine of Progress
Whereas progress in accelerated servers isn’t any shock, the market is rising quicker than anticipated, in line with Dell’Oro researchers.
“Accelerated server income did exceed our prior forecast or expectations,” Fung mentioned. “Momentum for accelerated server adoption is powerful, particularly for the hyperscale market.”
GPU-accelerated servers are outpacing regular server progress within the information heart by way of each income and shipments.
Fung mentioned that over 5 years from 2023 to 2028, the compound annual progress price for general-purpose servers is forecast to develop by solely 3%, versus 31% for accelerated servers.
AI workloads are forecast to develop at a tempo of 25-35% by means of 2027, in line with Bain & Firm’s newest World Know-how Report.
“Generative AI is the prime mover of the present wave of change, however it’s difficult by post-globalization shifts and the necessity to adapt enterprise processes to ship worth,” mentioned David Crawford, chairman of Bain’s international know-how apply.
Accelerated Computing’s Impression on Data Center Capex
The affect of accelerated computing on information heart capex is being felt around the globe, nevertheless it gained’t final without end, analysts mentioned.
“It’s each growing as a share of the entire and doubtless additionally driving the entire market measurement,” Omdia analyst Alexander Harrowell, instructed Data Center Information.
“That mentioned, it’s growing considerably quicker than the entire market, so sooner or later, the basic S-curve dynamic will kick in, in all probability within the subsequent two to 3 years, simply because the pool of potential adopters will probably be exhausted,” Harrowell mentioned. “That may depart us with a world of way more heterogeneous computing.”
When it comes to market dynamics, Harrowell mentioned Nvidia’s information heart enterprise is more likely to be round 10 occasions the scale of Intel’s information heart and AI division this 12 months. The market is additional buoyed by hyperscalers’ customized ASIC chips, together with choices from chipmakers together with AMD and Huawei.
AI Server Market Leaders and Income Distribution
Within the aggressive panorama of information heart server suppliers, Dell emerged because the chief by income for Q2 2024, adopted by HPE and Supermicro, in line with Dell’Oro evaluation.
Notably, accelerated servers accounted for an estimated 41% of OEM server revenues in the course of the quarter, highlighting their important affect available on the market.
The elevated demand for accelerator providers can even seemingly drive extra spending in associated areas as properly, akin to networking.
Fung famous that networking can also be benefiting from a rise in accelerated computing adoption.
“Accelerated servers normally require extra community bandwidth than general-purpose servers,” he mentioned.
“If the accelerated servers are interconnected in a big cluster that’s used for the coaching of huge language fashions or generative AI fashions, then a devoted community, known as a back-end community, could be wanted.
“This back-end community is even quicker than the front-end community, and there have been a variety of current improvements behind this new market.”
