The State of Media Infrastructure: The Year the Hardware Boxes Lost

For decades, broadcast ran on boxes. Expensive, single-purpose hardware, racked and wired, each doing one job very well. In 2026, that era is ending in earnest. The boxes are losing to software, and the shift is reshaping how media gets made and moved from the cable up.

IP is no longer the plan, it’s the deadline

Moving to IP-based workflows used to be a roadmap item, the thing everyone agreed they’d get to eventually. Two forces turned “eventually” into “now”: aging hardware that’s expensive to keep alive, and the ongoing reallocation of C-band satellite spectrum, which is squeezing the transponder capacity broadcasters have leaned on for decades. The FCC’s Upper C-band proceeding (3.98–4.2 GHz), published in December 2025, is the clearest signal yet that the ground is moving under satellite distribution. Broadcasters no longer have the luxury of waiting.

Most aren’t ripping everything out overnight. They’re running hybrids, bridging legacy SDI gear to an IP core so the migration happens in stages. But the direction is settled. Research from Caretta and Zixi found broadcasters are rebuilding live operations around IP as satellite and dedicated fibre decline, and it’s not theoretical: Sky and the NHL already run IP contribution feeds and cloud-based master control to cover more simultaneous events, add camera angles, and cut highlights faster than a truck full of hardware ever could. Software-defined workflows replace the custom boxes, trading longterm satellite capacity commitments for something that scales up and down on demand.

The catch is people. A broadcast engineer who mastered SDI didn’t necessarily sign up to become a network architect, and the skills to run software-defined systems live in IT, not the machine room. That gap between the expertise teams have and the expertise these systems need is the quiet crisis of this transition.

AI as the wiring, not the show

The interesting thing about AI in media infrastructure is where it isn’t: it’s not off generating the creative. It’s underneath, doing the unglamorous connective work that ties production, distribution, and archive together.

Mostly that means metadata. AI engines watch the video and listen to the audio, then tag what they find, objects, scenes, dialogue, so content stops being a mystery file nobody can search. It also means coordination: agents shepherding assets across platforms and firing off the next step without a human babysitting each handoff. And after the legal precedents set in 2025, there’s a hard preference for “authorized AI,” models trained on licensed, owned data so the chain of title holds up and nobody gets sued over training material.

Cloud came back down to earth

The cloud story matured from “put everything up there” to something more pragmatic. Hybrid won. Public cloud handles the elastic, scale-out work; on-prem and edge handle the things where latency and cost control matter, like live sports. Containers and Kubernetes are just the default building blocks now. And to dodge power limits and supply-chain risk, media shops stopped betting solely on one GPU vendor, spreading across AMD, Intel, and ARM alongside NVIDIA.

5G broadcast moved from trials to real deployments, with private stadium networks powering things like referee cameras and live player tracking at the 2026 World Cup. Underneath it all, data itself became the product: content without rich metadata is now treated as a liability, expensive to store and hard to monetize.

Where we fit

Read the whole shift in one line: media infrastructure is becoming software, and software is a different discipline than hardware. That’s opportunity and hazard in equal measure. The teams that thrive are the ones who can bridge the old craft and the new architecture without dropping the signal in between.

That bridge is the work we do. Helping broadcast and production teams move to IP, fold AI into operations without losing the human judgment that makes the output good, and do it with playbooks drawn from tools we run ourselves. As NVIDIA’s Richard Kerris put it, “the future of media production lies at the intersection of AI innovation and robust networking infrastructure.” We’d add one thing: someone has to stand at that intersection and actually direct traffic.


References: Broadcast in 2026: Nine Trends and Predictions, TV Technology; Caretta Research: Broadcasters Rebuild Live Operations Around IP, Zixi; In the Matter of Upper C-band (3.98–4.2 GHz), Federal Register.