- Telcos aiming to modernize their IT stacks for the AI era are running into a global memory shortage
- Instead of refreshing hardware, operators are increasingly trying to “sweat” existing servers, Red Hat’s Fran Heeran said
- Telcos and enterprises alike are also facing AI decision paralysis
Telcos the world over are looking to modernize their IT stacks to keep up in the AI era. There’s just one big problem: a global memory shortage has completely blown up the budget and left operators looking for a plan B.
“We are seeing telcos looking to get more out of their existing assets and maybe extend the lifecycle of those assets for longer than maybe they would normally do,” Red Hat telco chief Fran Heeran told Fierce.
Why? As Fierce has noted before, memory is in short supply. That, coupled with increased demand from AI players, has driven server costs up dramatically, in some cases tripling prices, Heeran said. So, telcos who had been planning to refresh their hardware are now looking to squeeze more life out of the compute they already have to keep modernizing their applications.
Heeran told Fierce this is one of the top three problems operators discussed during a roundtable at the Red Hat Summit last week.
“It's [about] not allowing the supply chain on servers to be a choke point on the ongoing transformations inside the carriers,” Heeran said. “The choice that is being made is, ‘I'm going to keep my existing servers. Now, who can help me sweat more out of those assets – get more lifetime out of them but still allow me to continue on my modernization journey in getting my VMs onto the right platform, getting my containers and VMs under single management?’”
That, of course, is where Red Hat comes in. Many telcos have already tapped the vendor to modernize their core networks with cloud-native containers, but its common cloud architecture can also be used to manage the virtual machines on the IT side of the house.
Heeran said the play is all about operational efficiency and collapsing siloes to eliminate redundant effort and compute capacity that is today dedicated to lifecycle management.
“Where you have multiple silos in place, you're managing all those silos separately,” he said. “If you can look at things through a single pane of glass across both your IT and your core network, across your VMs and your containers, if you can do all of that but also packing them into a more optimized set of compute, right?” That’s the goal.
AI paralysis
The pace of AI change is also creating a bit of progress paralysis among telcos, Heeran said.
Gone are the multi-year silicon lifecycles, replaced by new GPU drops every six to 12 months. That has operators questioning how future-proof the AI investments they may make today will really be.
The big question AI players are grappling with now is whether they’re moving too fast, to the point where “everyone’s going to say, ‘Well, I’m going to wait to invest until things start to slow down,’” Heeran said.
He added that the question was a universal conversation he had with customers at last week’s summit.