Google, for its half, will proceed to be “one of many largest, if not the most important,” client of compute assets, stated Bill Wong, analysis fellow at Data-Tech Analysis Group.
“Its enterprise mannequin drives that world demand, particularly by means of the widespread use of Google search and Gemini, which it gives for ‘free,’” he identified. Nevertheless, that very same stage of traction for enterprise clients is unlikely, as each Microsoft Azure and Amazon AWS have stronger footprints in enterprise.
AI infrastructure can be being influenced by the rising pattern towards sovereign AI, the place the popular AI stack is extra regionally managed or on-premises, Wong identified. International locations like Denmark wish to migrate each AI and non-AI workloads away from US suppliers, significantly Microsoft and Google.
However let’s see what inferencing brings
It’s additionally vital to notice that these numbers largely replicate infrastructure buildouts focused at large-scale coaching, a realm that Nvidia has dominated with its chips and its CUDA parallel computing platform.
However market share will seemingly shift as inference begins to mature, Kimball predicted. Suppliers like AMD and Cerebras will start to achieve as a result of they’re “equally spectacular,” and have totally different value and efficiency profiles, he stated.
The rankings additionally don’t account for some customized accelerators, together with AWS’ Trainium, Microsoft’s Maia, and Meta’s MTIA. Cloud suppliers will seemingly deploy their very own silicon “at any time when and wherever attainable,” as a result of there will likely be appreciable value and efficiency benefits, Kimball identified.
