HPE's 3-year telco AI plan

  • An HPE SVP told us the telco AI opportunity is at hand

  • He expects the next 3 years will be key

  • Will the slowcoach telcos keep pace?

HPE wants to be part of telco artificial intelligence (AI) infrastructure picture, the company's SVP and GM of telco, OEM and service providers Phil Cutrone told Silverlinings in a discussion at Mobile World Congress (MWC).

Cutrone said that he sees no bumps ahead for HPE’s forthcoming $14 billion acquisition of Juniper Networks and is not expecting regulatory trouble in China or anything of that sort. “Unless something crazy were to happen then we don’t see any reason why it would not move forward,” Cutrone commented.

This will be important going forward because Juniper is the more advanced AI partner out of HPE and HPE's Aruba brand.

Cutrone sees Juniper's AI smarts going towards enterprise infrastructure initially but notes that telco AI is just around the corner.

The path to telco AI

“When I look at the telco space I break it into two,” Cutrone said, when talking about AI. “That first part is the network and AI,” he added.

This could include AI and machine learning helping subscriber management, optimizing hand-offs between towers and automating the 5G core. Self-optimization and further automation will be key on the network side, Cutrone noted.

“Part two of this is monetization of the network, which is really key,” Cutrone said. He stated that network operators spend billions on spectrum and infrastructure but they want to monetize these building blocks in the way that hyperscalers and SASE players have. AI represents a real opportunity to do that.

How exactly, we wondered?

“The opportunity for them is to get in at the right time, to have the infrastructure needed for tuning for enterprises and for inferencing at the edge,” Cutrone said. “Since [telcos] have [a] backhaul network to do it and they have the data center locations...they are in a very advantageous position to capitalize on that. We are working very closely with them now to make it real,” he added.

The GM said that telcos are starting to “sketch out” their requirements right now. He noted that they want “line of sight” – presumably, to enterprise customers spending money on AI improvements – because the infrastructure spend isn’t going to be cheap (surprise!). Even GPUs are scarce and expensive right now.

Cutrone, however, expects – that despite their reputation for being slow and steady – telcos might actually move fast on deploying AI infrastructure.

“I think we’re looking at between a 12 to 18 month horizon for them to start deploying infrastructure to do this,” he stated.

Cutrone suggests that by 2025 we’ll “start to see the reality of AI” with initial investments incurred and some of the first use cases for telco AI. “The year after that” [which would be 2026, folks] “you’re going to see some of the actual use cases deployed and start to be monetized,” he said.

So, that might seem like a marathon for any other company, but for telcos that will be a sprint! Will they get it done? We’ll just have to wait and see in this case!