… and networks are on the hook. What has caught the eye today, news-wise? Well, it is slower. But even if the plug has been pulled on the MWC party, and the noisy future-gazing has quietened down, the market never shuts up. AI is still raising the roof. There is good stuff out there, and ways to continue the conversation RCR has been having with the industry for some months, at MWC as well. So a couple of things: Equinix, with 270 data centers in 77 metro markets, has introduced something called a Distributed AI Hub. Which is on our mind because we spoke to the company yesterday, and because what it said was interesting, and confirmed the recent RCR narrative about AI networks for AI workloads: that connectivity is front and center. 

Equinix’s hub proposition shows why: enterprises aren’t just moving workloads anymore, they’re trying to stitch together training in one cloud, inference in another, and agentic AI at the edge – all without breaking performance or latency budgets. What stands out (where Equinix hangs its hat, what it talked about yesterday) is the need for neutral facilities, standardized footprints, virtualized stacks – and dense and programmable fiber backbone networks. See coverage of Verizon, Cisco, Nokia, Lumen, Ciena – of late; plus all the stuff that came out of PTC two months ago (search the archive). Equinix wants to make it easier to place workloads where they make sense – avoiding egress costs, keeping traffic in jurisdiction, and connecting seamlessly to public clouds, private clouds, neo‑clouds. 

Its $4 billion acquisition of atNorth strengthens its edge footprint, also, while integrations like Prisma Air add security and compliance guardrails for increasingly complex AI traffic. But the networking piece is just as crucial, and an urgent AI focus for the firm. Its ‘fabric’ (Equinix Fabric) and last-mile (as-a-service) plays simplify connections between enterprise sites, clouds, and AI partners. And actually, all of this chimes with what’s going on at the ‘far edge’, on enterprise premises with private networks – what RCR has always written about. It is the same table stakes: high-performance, total reliability, tight security. Check out Ericsson’s smart moves to mop-up Nokia’s silly mess in private 5G, and ramp it up for private 5G. 

Point is: data centers are the engine rooms for the AI economy, but they are teetering under their workloads, and being distributed far and wide in search of power, and in a quest to serve global markets, and satisfy compliance and regulation. The challenge is to connect them, and as inference loads ramp up over the next few years, the east-west interconnect focus will turn north-south, as AI goes to work at the metro and the enterprise edge. Same rules apply: the network has to be performant, programmable, predictable. Happy days for smart telcos, surely? Or potentially, anyway.