Padraig O’Donovan’s story about turning broadcast kit into infrastructure code sounds brilliant — until you ask who owns the workflows when the cloud bill arrives.

By Ben Anchor — Wednesday, 1 April 2026 · Listen to the podcast episode

I’ve been mulling over something Padraig O’Donovan said on the podcast this week. LayerCake’s pitch is compelling on paper: “We can turn on an end-to-end broadcast production solution in the cloud in about half an hour. That would have been unheard of — the kind of solution that would have taken six to twelve months to build in a studio once upon a time.” Half an hour versus a year. It’s the sort of claim that makes people sit up, myself included.

But I keep circling back to a question Padraig didn’t answer, perhaps because I didn’t push him on it: who owns the workflows when the cloud bill arrives?

The Allure of Deploy-and-Destroy

Padraig’s background — five years at a tier-one Australian broadcaster, leading a $7.5 million digital transformation at Channel 10, fourteen suppliers stitched together into one unwieldy ecosystem — gives him the credibility to say something uncomfortable: broadcast infrastructure used to be a nightmare. Hardware-based, studio-bound, months of cabling and integration before you could go live. LayerCake’s model offers the opposite: deploy for the event, destroy when you’re done, pay only for what you use. Cloud as the antidote to capital expenditure.

It’s seductive, and I don’t doubt the technical achievement. But I’ve sat through enough cloud migrations to know that the promise of elasticity often collides with the reality of operational accountability. When Padraig talks about building workflows that align infrastructure tiers with content value — top-tier content on premium infrastructure, lower-value stuff on cheaper compute — he’s describing a world where broadcasters treat content like a portfolio of assets with varying risk profiles. That’s clever. It’s also expensive if you get the tiers wrong, or if your event overruns, or if your destroy script fails and you’re left with a hundred instances chewing through AWS credits while everyone’s on the weekend.

The bit that stuck with me was Padraig’s comment about redundancy. He mentioned Cloudflare’s outages last year and said LayerCake is building solutions that let digital systems switch CDNs mid-stream, bringing the kind of failover traditional broadcasters had in hardware. That’s genuinely useful. But it also underscores a broader point: the cloud hasn’t eliminated operational risk, it’s just redistributed it. You’re no longer locked into a single studio vendor, but you are dependent on infrastructure providers, orchestration platforms, and the scripts that glue them together. If your automation fails, you’re not dealing with a broken cable — you’re dealing with a broken workflow that might be distributed across three regions and five suppliers.

I asked him about the Australian market, partly because I wanted to understand whether the trends we’re seeing in the UK — YouTube as the number-two TV platform, streamers carving off younger audiences — are playing out the same way down under. They are. Padraig’s observation that “the traditional TV audience is declining” and broadcasters are scrambling to “commercialise content in many different formats” is familiar territory. What caught my attention was his story about Clive Dickens at Channel 7, who convinced the network to live-stream the Olympics and then left the live streams running afterwards. That decision — keeping linear channels live in digital form — turned out to be commercially transformative, because it opened up new ad inventory. It also validated the idea that audiences want optionality: live if they’re free, catch-up if they’re not.

AI as the Next Orchestration Layer

Padraig’s most bullish about AI, and I can see why. He rattled off use cases: Magnify pulling highlight clips from live streams using AI, pushing them to social with brand integration; iSports using AI to analyse sports vision in real-time; SportsVisio processing grassroots volleyball and basketball to extract insights. These aren’t speculative — they’re happening now, and they’re turning content that would’ve been ignored into monetisable assets.

But here’s where I get twitchy. AI in broadcast workflows is another layer of orchestration, another dependency. Padraig described LayerCake’s StreamCake platform as letting you “switch out” a component — say, a commercial workflow tool — without impacting the rest of your pipeline. That’s great if the component you’re swapping is a known quantity with stable APIs. But AI models evolve, they drift, they require retraining. If you’re building a workflow that depends on an AI partner’s black-box analysis of live sports, you’re trusting that their model won’t degrade, that their latency won’t creep up, that their pricing won’t pivot when you’re locked in.

I’m not saying don’t do it. I’m saying the shift from “fourteen suppliers stitched together” to “agnostic orchestration platform with interchangeable components” sounds cleaner on the surface, but you’ve still got dependencies — they’re just more abstract now. The hardware studio had the virtue of being tangible. Cloud workflows are composable, yes, but they’re also opaque unless you’ve got serious observability tooling and someone who understands the entire stack. How many tier-one broadcasters have that?

What Happens When the Savings Evaporate?

Padraig claimed LayerCake’s model can save “up to 90%” on infrastructure by letting clients choose providers and align costs with content value. I’ll take that at face value for now, but I’d want to see the assumptions. Ninety per cent savings compared to what baseline? On-prem hardware with full redundancy? A naive cloud deployment where everything runs on premium compute? And crucially: are those savings sustained over time, or do they erode as workflows grow more complex, as you add AI partners, as you scale for live events that don’t behave like your projections?

The promise of cloud has always been elasticity, but elasticity cuts both ways. You can scale down, sure — if you remember to destroy what you spun up. If your automation works. If your commercial contracts don’t lock you into minimum commits. I’ve seen too many cloud bills balloon because nobody was watching the usage dashboards, or because “deploy and destroy” turned into “deploy and forget.”

I don’t doubt Padraig’s technical chops or LayerCake’s capability. The platform sounds genuinely innovative, and the partners he’s working with — Grass Valley’s GVAM, Magnify, SportsVisio — are serious names. But the underlying question is one of ownership. Who owns the orchestration logic when your infrastructure is distributed across AWS, Oracle, Google, and Linode? Who owns the runbook when a CDN fails mid-stream? Who owns the model drift when your AI partner’s accuracy slips by five per cent?

Cloud production doesn’t solve the problem of complexity. It just moves it from the hardware layer to the software layer, and from the studio floor to the dashboard. Whether that’s better depends entirely on whether you’ve got the people and processes to manage it. Most broadcasters, in my experience, don’t — yet.