GPUaaS represents a significant alternative — however ABI Analysis Analyst Larbi Belkhit mentioned that income isn’t the only real motivation behind telcos’ AI manufacturing facility buildouts
Telecom operators are positioning themselves as main gamers within the synthetic intelligence (AI) infrastructure house, with ABI Analysis projecting they may generate greater than $21 billion in income from GPU-as-a-Service (GPUaaS) by 2030.
“Nearly 100% of that is going to be AI inference within the workloads, not coaching,” mentioned ABI Analysis Analyst Larbi Belkhit, creator of a brand new report on the pattern, in an interview with RCR Wi-fi Information.
The report recognized roughly 17 telcos worldwide which are at the moment constructing AI factories utilizing Nvidia GPUs. The bulk are situated within the Asia-Pacific (APAC) area, with a number of underway in Europe, the Center East and Africa, and one introduced in North America.
APAC telcos are taking the lead, Belkhit defined, as a result of many are shifting away from conventional telecom companies to embrace AI as a core enterprise: “In our analysis, we confer with this technique as telcos turning into sovereign hyperscalers” — nationwide or regional gamers that personal and function their very own AI infrastructure.
Standout operators within the area embrace Singtel, which introduced in March 2024 a partnership with Nvidia to democratize AI by GPUaaS. Different APAC leaders embrace SoftBank, SK Telecom and Korea Telecom. In Europe, Telenor is a front-runner, although Belkhit mentioned different European telcos are starting to noticeably consider the chance.
In North America, Verizon has taken a step on this route with the launch of Verizon AI Join, a collaboration with GPUaaS and cloud computing supplier Vultr. The platform permits enterprises to deploy AI workloads at scale. Nonetheless, Belkhit famous, “They don’t have a direct play but — they’re extra of a facilitator.”
So, how will hyperscalers react to telcos shifting into GPUaaS? They in all probability don’t see them as a lot of a menace, mentioned Belkhit — at the least not proper now. “In the previous couple of years, the newer cloud suppliers or neoclouds have been gaining a number of traction.” He added that telcos are partnering with these neocloud gamers, which may pose a extra strategic problem to the hyperscalers. “From a hyperscaler’s perspective, it’s about coping with these neocloud firms first … then maybe telcos companion with the hyperscalers as a substitute.”
GPUaaS — not precedence primary, and never the ultimate step
Whereas GPUaaS represents a significant enterprise alternative, Belkhit clarified that income isn’t the only real motivation behind telcos’ AI manufacturing facility buildouts. “They’re firstly constructed for inside workflow functions and integrating AI,” he mentioned. “The GPU as a service alternative …. is an entryway into the AI marketplace for telcos [and] not the ultimate step … as a result of it goes again to what telcos are recognized for — managing infrastructure.”