MWC 2025: Fabric to spread as BT sees no business plan B without the network

BT has said it is seeing technical breakthroughs translating into massive opportunities for service providers, and one of its key customers in this arena, Colin Bannon, CTO of BT Business, told a Cisco event at MWC 2025 that in this reshaped world of work, the network is the computer and is a fundamental prerequisite for artificial intelligence (AI) to work.

The presentation in Barcelona came just days after BT announced customer traffic was now live on its AI-ready Global Fabric network-as-a-service (NaaS) platform, which is designed to connect the multiple clouds businesses use for their applications and data and will allow them to take advantage of the new wave of digital automation and AI.

Running on the underlay of the deterministic network is an AI-powered digital orchestration layer, with BT supplying customers with a predictable application experience they expect by selecting the optimal end-to-end paths for their applications and workloads as they move to and between multiple clouds and end users.

Options will include a BT-enhanced internet service, point-to-point Ethernet, multi-point Ethernet and MPLS, and these will be offered in bandwidth increments of 1 Mbps up to 100 Gbps. Connectivity will be interchangeable on the same port, offering flexibility that BT said is not possible on current networks.

When fully built out, Global Fabric will be available to customers internationally via 140 points of presence, hosted in cloud locations across 40 countries. It will offer 74% direct coverage of hyper-scaler clouds and pre-provisioned, high-bandwidth connectivity to more than 700 datacentres.

Assessing the opportunities that the new infrastructure could provide, Bannon observed that how to stay relevant for the ongoing challenges faced by the customers BT serves, mainly from international business, governments and multinationals, is a major challenge. The company, he believed, needed to rethink how to address business customer priorities in a world of increasingly distributed workloads and progressively sophisticated cyber threats.

Even though many opportunities existed through the new platform, Bannon was adamant that trust was one thing that BT simply could not give up on and that network performance to support the new AI enabled was also crucial in a world where “slow is the new down”.

He added: “You’ve got to where you have this journey where you need full stack observability of the [network] experience and of the applications going on there. We used to design networks [that were] persona-based [looking at] people, things and apps. Now we have agents as well, from an AI perspective, and all of them have a set of requirements on what they’re doing.

“So, providing a service where you deliver outcomes on both assurance and security but also end user experience is going to be really important. [And we need to think] about our mobile network and the convergence of the fixed network and those outcomes.”

Enterprises are thinking very closely about mobile networks – in particular, 5G networks carrying AI workloads. A key element is the rapidly increasing amount of AI that is being loaded onto handset as manufacturers begin moving AI models on the devices, and this is seeing IT leaders question company policies reading as to exactly where data generated though AI services is stored and how secure it is.

“Our goal is to have the best network for AI in the UK for mobile as well as fixed. It’s really ‘watch this space’ for the enterprises,” said Bannon. “It is interesting. The vendors are turning [AI models] onto the devices, but I speak to a lot of CIOs, and six months ago, about half of them said, ‘I’m thinking about stopping bring-your-own-device [BYOD] policies’, because they’re afraid of data loss [such as from] new employees, new generations who are very technologically savvy [who are] installing dictation AI apps on their BYOD devices, [and companies don’t] know where the data is necessarily going with the third-party services.

“I think having overlay services, AI defence and data loss protection to deal with that [is preferable to] banning things. In the next 12 months, AI will be the new UI for these mobile devices. We need to work with the manufacturers to get a really good capacity model and understanding of the fingerprint of the network to [find] the propensity model of AI traffic on mobile networks. That way, we’re able to make the adjustments in the network and be able to serve it brilliantly. That’ll be a big piece of work as we go forward to make sure that our mobile network is AI ready.”

Bannon noted that there would be number of key elements on which BT, working with Cisco, would need to develop in order to realise the full stack of Global Fabric. This involved the SD-WAN and the data sovereignty, the latter case not just for data at rest in storage but also for data in motion. If you have standard mode between two ISPs on the internet, and data reconverges, it may reconverge outside of the hoist country.

“The internet is brilliant at recovering data; it was designed to survive a nuclear war, originally. However, it is geographically ignorant. Do you need to have an additional layer of intelligence on the overlay or the underlay? You can’t fix that if you’re on an SD-WAN if the SD-WAN is going underneath. You can’t geofence in that country. But if you have that probability under the underlay, we can start to geofence failure modes to keep the data flows sovereign, not just the data residency in the datacentres.

“My belief is that you will get better outcomes with both a programmable underlay and a programmable overlay with full stack of observability, and the two working in sympathy with the managed service around it to deliver the best outcomes and the best operational availability. But if you think about how people use [the network] with people being mobile now, having that mobile overlay capability and [offering] zero trust is still important.”

What will likely be just as important is having to construct enterprise networks that can cope with the massive strain placed on utilisation from AI, especially in terms of what is needed for training models. Bannon said that Global Fabric was constructed before AI “blew up” and that it was “fortuitous” that BT was thinking of a hyper distributed set of workloads on a network and to be able to make it fully programmable, “where you’d abstracted the control plane from the data plane”, which it hadn’t done before. Every router in its core used to compute the path to the next column, and “you didn’t have that control”, he added.

Looking at the challenges ahead and how the new network will resolve them, he concluded: “[AI] just makes distributed and more complex workflows even bigger, which makes the need for a fabric type network even more important. So, in the timing, there’s always a bit of luck.

“You need a network that can burst and that is programmable and that you can do bandwidth on demand as well. All of this programmability you’ve never had before. Sometimes it’s being in the right place at the right time. I would argue that the network is the computer and the network is a prerequisite for AI to work.”