Ray Le Maistre, TelecomTV (00:13):
Telcos all over the world are currently grappling with their AI strategies, and there's no shortage of potential partner companies willing to help them out. And one of those companies is HPE, and I'm talking today with Martin Halstead, senior distinguished Technologist, Aruba Telco Solutions at Hewlett Packard Enterprise to find out how the giant tech company is helping network operators with their AI plans. Martin, thanks so much for joining us today. So we've heard quite a lot from HPE lately talking about various AI activities. Can you share HPE's overall AI strategy and how it applies to the telco sector?

Martin Halstead, HPE (00:54):
Sure. So at HPE, we see the AI industry as being, it's a 35 billion market. It's growing at an annual rate of 25%. But the interesting part for us is that 80% of those AI workloads are based on the customer premises. So from a telecom's perspective, that's a very interesting place for us to commit a significant amount of resource and effort in addressing. Now what is our legacy in telecoms? And it comes really from our years of deployments in some of the largest telecoms networks in the world through our pro lines portfolio. But when we couple that alongside our super computing platform, we've built some of the largest and most complex supercomputers in the world, specifically for handling AI for a number of use cases. Made a number of announcements in that space at HP Discover in Las Vegas this year alongside Nvidia. So our strategy for telecoms is to leverage the development works that we have for other industry verticals, but equally apply that to the telecom space.

Ray Le Maistre, TelecomTV (02:24):
And what does HPE see as the top AI use cases for telcos, and how is HPE enabling the development, deployment and management of those solutions?

Martin Halstead, HPE (02:37):
Well, we see the development of AI specific use cases in telecoms being split into three distinct areas. There is the customer facing and resource facing applications that the telecoms operators use. So these are applications that do the cross domain provisioning or an orchestration of solution sets. So for example, network slicing could be one use case for that, but also single domain orchestration as well. So the orchestration of network functions within a data center as a vertically integrated stack and the use of AI in those types of use cases. So the second one is network planning and dimensioning for that. It is all about how do you create a digital twin of the telco domain that you are attempting to replicate. And then once you have built that digital twin, you can look at how the best times for doing things like maintenance, windows, upgrades, et cetera.

(03:54):
The third area then is network performance and user experience. And as we expand service offerings in telecoms across multiple domains, the amount of different data types that are ingested in order to understand whether the user experience is sufficient to meet the KPIs of the operator becomes more and more complex. And so AI and in particular, machine learning can help with that. So at HPE, we have a dedicated business unit that develops operational support system, OSS software, and that third area of machine learning and moving into generative AI for assurance type use cases is an area that we have under development today.

Ray Le Maistre, TelecomTV (04:49):
So a lot of key software developments there. But how is HPE broadening its HPE compute portfolio to enable more consistent and general purpose infrastructure to enable more open and standards-based AI solutions?

Martin Halstead, HPE (05:05):
HPE has a number of innovations in the supercomputing area, principally through our Cray portfolio. And we made a number of announcements at HPE Discover in Las Vegas this year alongside Nvidia. But when it comes to proli and the deployment legacy that we have within the telecoms environment, we see the edges for AI being incredibly important. And so to that end, we have specific AI based developments in terms of two product sets. The DL 3 84 2 Rack unit, GPU enabled server developed alongside Nvidia and AMD as well as the DL three 20 A one rack unit server. But also with that and kind of cognizant that the telecoms operators don't necessarily have the skill sets to pull together a full stack solution for AI and build that cost effectively. We have also, in addition to our usual CapEx based sales motion into telecoms operators, we have HP GreenLake where AI can be consumed as an offering. And so with that, we have GreenLake private cloud AI, where we have different T-shirt sizes of AI infrastructure that can be used principally for inferencing and again, have been developed alongside nvidia.

Ray Le Maistre, TelecomTV (06:49):
Okay, that's a good range of options there. So what's next for AI in telecom and how is HPE tackling some of the challenges that telcos will face in 2025 and beyond?

Martin Halstead, HPE (07:05):
So the advantage that we see for HPE in telecoms environments is that we have a full stack solution. So we have a full portfolio that can handle the intricacies of AI from the infrastructure itself being our super compute platforms in the form of Cray, as well as Proliance servers and the additions that we are building for them. We also have an ML ops environment from our own software developments as well as those coupled with NVIDIA and then our OSS portfolio where we can have applications developed, which would help address the TCO questions that will principally come up from the operators. So from our perspective, we see that our involvement in AI in telecoms will lead the telecoms operators who are typically very risk averse in moving to ai. We can help them in that journey.

Ray Le Maistre, TelecomTV (08:24):
So clearly a lot there in the HPE portfolio as telcos address AI strategies. Martin, thanks very much for giving us the overall picture there of what HPE is doing and the trends in the industry. Really interesting stuff and thanks for joining us today.

Martin Halstead, HPE (08:42):
Thank you.