Analysis: The industry quietly changed what it wants from the cloud

Two recent Industry Insights roundtable series — one on cloud production, the other on workflows and integration — covered different ground but kept arriving at the same conclusion. The question broadcasters asked five years ago, whether to move to the cloud, has largely been answered.

The harder questions now concern how to run distributed operations without losing the consistency that fixed facilities once guaranteed.

Read together, the two conversations trace a shift in priorities. Flexibility, for years the selling point of cloud and IP-based production, is no longer the headline. Operational predictability is.

Flexibility gave way to predictability

For much of the past decade, vendors pitched the cloud on elasticity: spin up capacity for a live event, tear it down afterward, pay for what gets used. That argument still holds, but participants in both series framed it as table stakes rather than the goal.

“The industry talked about flexibility almost as if that alone was the goal,” said Yaya Selva, CMO at Net Insight. “Flexibility matters, but not if the behavior of the live service becomes harder to predict when the pressure is on.”

That sentiment ran through both discussions.

Rene van Koll of Big Blue Marble noted that cloud workflows scale quickly but require more structured design to behave consistently, with the predictability itself coming from established broadcast engineering practices rather than the infrastructure.

The reason is the nature of live production, where a dropped frame or a timing error reaches viewers immediately. Latency, the delay between a signal entering a system and leaving it, remains the constraint that keeps some functions tethered to dedicated hardware.

Advertisement

“Latency is still one of the biggest considerations, especially for live production where timing is everything,” said John Mailhot, SVP of product management at Imagine Communications. Most organizations, he added, keep latency-sensitive parts of the workflow close to operators while using the cloud elsewhere in the pipeline.

Hybrid stopped being a phase

If flexibility was the old promise, hybrid is the settled answer.

Across both roundtables, participants described hybrid architecture, a mix of on-premises hardware, edge processing and cloud resources, not as a stepping stone toward full cloud adoption but as the destination.

“Hybrid isn’t a transitional state. For most broadcasters, it’s the end state,” said Golan Simani, director of cloud and technical operations at TAG Video Systems. The economics and risk profile of a full migration, Simani noted, do not work for everyone, and many organizations will run both models in parallel for years.

That framing recasts a debate the industry has had since IP infrastructure became viable. Rather than cloud versus facility, participants described a single workflow spanning both. Ian Wagdin of Appear put the principle plainly: keep the parts that benefit from deterministic performance in dedicated infrastructure and use software and cloud where elastic scale adds the most value.

The practical appeal is cost predictability. Broadcasters keep high-utilization workflows on-premises, where capital investment produces a known cost, and reserve the cloud for overflow, resilience and short-term needs.

Orchestration became the connective layer

The word that appeared most often across both series was orchestration, the software layer that coordinates routing, automation, content management and monitoring across systems that were never designed to work together.

In the cloud production discussion, orchestration was the mechanism that made distributed teams workable. In the workflows series, it was the thing turning a collection of tools into something resembling a supply chain.

“The biggest shift has been the move to orchestration as a central control layer,” said Mailhot. Once it is in place, he noted, operators can manage services across on-premises and cloud environments without needing to know where anything physically runs.

Underneath orchestration sits a quieter agreement about open standards. Several participants pointed to SMPTE ST 2110, the suite that carries video, audio and data as separate IP streams, as the common transport layer that lets equipment from different vendors share one workflow. The consensus on the alternative was blunt.

“The era of the single-vendor end-to-end solution is functionally over,” said Rich Zabel, VP of media supply chain at Diversified.

Well-maintained APIs, Zabel noted, have become a procurement requirement rather than a feature to evaluate, because a best-of-breed system only works when components can be swapped without rebuilding the pipeline around them.

Advertisement

The supply chain grew up

The phrase “media supply chain,” borrowing manufacturing’s vocabulary of throughput, cycle time and cost per unit, recurred enough to suggest it has moved past metaphor.

“The media supply chain stopped being theoretical a few years ago,” said Geoff Stedman, CMO at SDVI. The open question, he noted, was never whether the framework was useful but whether the technology existed to implement it at scale.

What changed, by several accounts, was automation.

Tasks that once required human attention at every stage — ingest, quality control, transcoding and metadata tagging — are increasingly handled by software, with operators stepping in for exceptions rather than working through every file.

Zabel framed the stakes in terms that will resonate with anyone running a lean team: automation has moved from a competitive advantage to a survival requirement, because a single pipeline now feeds broadcast, streaming, social and syndication at once, and a bad file no longer creates one problem but a cascade across every delivery point.

Metadata is the connective material that makes this work.

“Metadata is king,” said Guillaume Aubuchon, VP of product management at Avid, describing a shift toward treating it as persistent and portable rather than bound to a single system. The recurring complaint was preservation: content passes through asset management, transcoding, quality control and distribution, and there is no guarantee the metadata generated at one step survives the handoff to the next.

What “mature” now means

The most telling answers came when participants were asked to define a mature cloud production environment.

Almost none pointed to how much of it sits in the cloud.

A mature environment, Selva noted, is one where operators can move live workflows across facilities, cloud, networks and partner domains while keeping behavior predictable and control intact. Wagdin offered a similar measure: how well the whole system operates, not how much of it has migrated.

Advertisement

Aubuchon went further, suggesting the mature environment would eventually be invisible to the people using it, with teams no longer thinking in terms of cloud versus on-premises at all.

There is a tension worth naming here. The same conversations that celebrated software-defined flexibility also returned, repeatedly, to discipline: financial operations practices, governance, monitoring and codified process. Simani’s warning that a $500,000 cloud budget can quietly become $1.5 million without cost controls was a reminder that elasticity runs in both directions, and that unmonitored infrastructure is expensive infrastructure.

The through-line, then, is not that the cloud won.

It is that the industry has stopped treating location as an interesting question. What matters now is whether a distributed operation behaves like a reliable one — and on that point, the vendors selling the tools and the engineers running them appeared, for once, to be in agreement.