The rising price of innovation: When hardware becomes a sunk cost in live production
When ownership defined capability
There was a time when owning the production kit was the hallmark of a serious live producer.
Racks of servers. Tower PCs running real-time graphics. Dedicated encoders. Graphics Processing Unit (GPU) cards stacked like trophies.
It signified capability, control, and technical confidence.
But the economics behind that model are shifting. Fast.As demand for powerful compute continues to grow, especially AI workloads, the components that make modern production possible are becoming not just advanced, but costly. And they are only getting more expensive, not less.
GPU costs are surging…and that matters
AI has reshaped how we think about live video production. It powers real-time clipping, automated highlight generation, smart encoding, speech-to-text, multi-angle reframing, and more.
All of these capabilities rely on advanced compute, typically delivered through high-performance GPUs.
But in the last couple of years the price dynamic around GPUs has changed significantly:
- Reports suggest manufacturers like NVIDIA and AMD are raising the retail prices of GPUs as demand for AI and high-memory parts has intensified.
- In some cases, prices for flagship GPUs have more than doubled compared with previous generations.
- Enterprise-grade AI GPUs can cost tens of thousands of dollars per card, with fully-configured multi-GPU systems stretching into the hundreds of thousands of dollars.
- At the same time, the high-bandwidth memory (HBM) and other components that power modern AI accelerators are in short supply, creating further upward pressure on pricing.
The net effect is clear. If you are trying to build and maintain your own hardware stack to support modern live production workflows, you’re now fighting a moving target where the price of entry keeps rising.
What is a GPU in live production?
In live production, a GPU is the hardware doing the real-time heavy lifting.
It processes, compresses, analyses and modifies video in milliseconds. When you’re encoding multiple streams, generating graphics, reframing vertically, detecting highlights or running speech-to-text, a GPU is powering it.
The more real-time manipulation and AI you add, the more GPU capacity you need.
And that’s where cost pressure begins.
Short upgrade cycles mean little return on investment
There’s a deeper economic challenge emerging across live production.
Broadcast hardware has traditionally held value over longer upgrade cycles. Equipment was procured, depreciated over time and often reused for years.
That model is breaking down in the era of heavy GPU-based production systems.
By the time organisations justify, procure and deploy the latest generation of hardware, the next iteration, offering greater memory capacity, higher compute and expanded AI capability, is already on the horizon. The impact is clear:
- Hardware is replaced more frequently
- Depreciation cycles compress
- Residual value erodes
- Capital is tied to infrastructure that quickly becomes outdated
In contrast, modern cloud-native services operate on continuous infrastructure evolution. Compute upgrades, performance gains and architectural improvements happen in the background, without disruptive replacement cycles.
Increasingly, conversations with CTOs reflect this shift. Innovation cycles are accelerating faster than traditional capital procurement cycles, creating structural tension for organisations still reliant on fixed hardware investments.
CapEx vs. OpEx – It’s not just about accounting
Shifting from capital expenditure (CapEx) to operational expenditure (OpEx) isn’t just a budgeting decision. It’s a strategic one.
When you own hardware, you are responsible for:
- Predicting future capacity needs
- Funding every refresh cycle
- Integrating new hardware into legacy stacks
- Maintaining physical infrastructure
- Managing depreciation and lifecycle replacement
With an operational model built on hyperscale infrastructure, the economics look different.
Cloud providers such as Amazon Web Services invest billions annually into next-generation compute environments. They absorb the cost of constant hardware refreshes and amortise that expense across thousands of customers.
For an individual broadcaster, league or independent producer, replicating that level of infrastructure investment on a single balance sheet is increasingly unrealistic.
A grounded example: Formula Drift
Formula Drift wanted to deliver more dynamic live content, vertical streams, social optimised clips, real-time distribution, without building and maintaining separate hardware paths for every destination.
Instead of investing in new racks of GPUs or constant cycle refreshes, they worked with Grabyo to leverage AWS’ Elemental Inference to leverage their infrastructure.
The result was a more flexible production workflow that could deliver multiple outputs, including live vertical formats, without the overhead and risk of owning the infrastructure itself.This use case isn’t about hardware vs. cloud rhetoric. It’s about where investment delivers strategic value for a live producer: into audience-facing experiences rather than capital refresh cycles.
The broader trend: ownership no longer defines capability
For decades, capability in live production was closely tied to ownership.
If you owned the infrastructure, you controlled the output. If you had the most advanced systems, you had the edge.
That equation is changing.
Modern live production is increasingly defined by speed, flexibility and distribution rather than physical assets. The ability to spin up new formats, launch vertical streams, deploy AI-driven features or distribute simultaneously across platforms matters more than the equipment sitting in a machine room.
Infrastructure is still necessary. But it is no longer the differentiator.
The differentiator is how quickly you can adapt.
Owning hardware does not inherently make you more innovative. In some cases, it can slow innovation down, because every evolution requires a new procurement decision, a new deployment cycle or another layer of integration.
When infrastructure becomes abstracted, innovation shifts up the stack.
The competitive advantage moves from “what do we own?” to “how quickly can we change?”
That is a fundamental repositioning of where value sits in live production.
Compute becomes assumed. Agility becomes strategic.
What this means for producers and broadcasters
Hardware still has value. But it’s shifting from a differentiator to a liability when it:
- Ages too quickly
- Costs more to refresh than it ever delivered in value
- Diverts engineering focus away from audience experience
Good producers are starting to ask themselves: “How do we invest in capabilities that outlive the next refresh cycle?”
That’s not a question about abandoning standards or quality. It’s about making the trade-off between sunk cost and continuous innovation.
In other words, invest in workflows that evolve, not just hardware that depreciates.
