The time period ‘AI information middle’ is showing with growing frequency in conversations about fashionable information facilities. That is unsurprising, on condition that fashionable AI know-how has positioned novel calls for on information facilities and the infrastructure inside them.
But, it could additionally really feel like a hole time period as a result of it stays unclear how, precisely, an AI information middle is totally different from different information facilities. Earlier than casually utilizing this terminology – or assuming that AI workloads require basically totally different information facilities than these already in existence – it is vital to make clear what the time period truly means.
What Is an AI Knowledge Heart?
The quick definition of ‘AI information middle’ is a facility that has been particularly designed for internet hosting AI workloads.
This isn’t an particularly satisfying definition, nonetheless, as a result of the wants of AI workloads can fluctuate broadly. Thus, there are not any distinctive information middle options that separate AI-oriented amenities from “common” information facilities.
Plus, few individuals had been speaking about AI information facilities till generative AI went mainstream with the launch of ChatGPT in 2022. That’s the purpose when the media began discussing AI information facilities as a novel sort of information middle. Earlier than then, different sorts of AI workloads – like predictive and descriptive analytics – had been working completely properly in conventional information facilities, and nobody was suggesting that AI required a basically new sort of information middle.
Reasonably than constructing specialised AI amenities, some organizations could discover higher worth in adapting present information middle infrastructure to help their evolving AI workloads. Picture: Alamy.
Key Options of AI Knowledge Facilities
That stated, it’s affordable to conclude that fashionable AI workloads do profit from information facilities designed particularly to fulfill their distinctive wants. Precisely how an information middle does this could fluctuate, however widespread options of AI-oriented information facilities could embrace:
-
Bigger general capability. Arguably, that is a very powerful defining attribute of an AI information middle: It’s merely greater – within the sense that it could present extra whole energy enter and accommodate a bigger variety of servers – than most conventional information facilities.
-
The supply of servers outfitted with GPUs or different {hardware} gadgets that may speed up some AI duties, like AI mannequin coaching and inference.
-
Optimized networking gear, like DPUs, that may assist to course of site visitors very quickly – one other vital requirement for a lot of AI workloads, particularly people who have to course of massive portions of information in actual time.
-
Excessive-efficiency information middle cooling techniques, like immersion cooling, that are vital in some instances for dissipating the massive portions of warmth produced by AI workloads.
AI Knowledge Facilities vs. Conventional Knowledge Facilities: A Blurry Line
The tough factor about making an attempt to outline AI information facilities primarily based on traits like these described above is that none of these options is exclusive to AI information facilities.
For instance, hyperscale information facilities – that means very massive amenities able to accommodating greater than 100 thousand servers in some instances – existed earlier than fashionable AI debuted. AI has made large-scale information facilities extra vital as a result of AI workloads require huge infrastructures, however it’s not as if nobody was constructing massive information facilities earlier than AI rose to prominence.
Likewise, it has lengthy been potential to deploy GPU-equipped servers in information facilities. AI has merely elevated the significance of entry to GPUs for sure use instances. The identical is true of DPUs, which have been round for years.
Likewise, superior cooling techniques and modern approaches to information middle energy administration will not be distinctive to the age of generative AI. They, too, predated AI information facilities.
What Is an AI Knowledge Heart, Actually?
Arguably, an AI information middle is finally outlined by what it does (internet hosting AI workloads) greater than by the way it does it (utilizing the identical sorts of IT infrastructure, energy and cooling techniques as another sort of information middle).
So, earlier than getting hung up on the concept that AI requires funding in a brand new era of information facilities, it’s maybe more healthy to consider methods to leverage the information facilities already in existence to help AI workloads. That perspective will assist the trade keep away from the danger of overinvesting in new information facilities designed particularly for AI – and as a bonus, it might get monetary savings by permitting companies to repurpose the information facilities they already personal to fulfill their AI wants as properly.
