NAB Show returns to Las Vegas this April, and with it comes the annual opportunity to take stock of where the industry is heading.
For Scalstrm, NAB 2026 represents something more specific: a moment to challenge assumptions that have quietly calcified into received wisdom, and to demonstrate, live on the floor, that there is a better way to run video infrastructure at scale.
The Old Model is Still Everywhere
Walk into most broadcast or OTT operations today, and you will find the same basic architecture: always-on transcoding pipelines, libraries of pre-encoded renditions stored across every conceivable bitrate and format, and racks of CPU-heavy infrastructure running regardless of whether a single viewer is watching. It is a model built for a different era, when encoding flexibility was limited and provisioning in advance was the only reliable option.
That model still works. But working is not the same as being efficient. The compute runs constantly. The storage multiplies. The energy costs accumulate. And when platforms try to scale, the cost scales with them, often faster.
The question the industry is increasingly asking is not whether this approach can continue, but for how long it can be justified.
“Streaming has scaled enormously. The infrastructure running underneath it, in many cases, has not kept up. At Scalstrm, we are focused on closing that gap, building a platform where high-quality, large-scale delivery does not come at the cost of operational complexity or unsustainable spend. The industry is ready for a better way, and we are here to deliver it.”
— Patrik Alm, Founder and CEO, Scalstrm
VPUs: Strong Potential, Misunderstood in Practice
Video Processing Units have generated significant attention over the past few years, and rightly so. Designed specifically for the data patterns of video encoding and decoding, VPUs deliver throughput and energy efficiency that general-purpose CPUs simply cannot match for high-volume, predictable workloads. The hardware case for VPUs is well established. What is less well understood is how to realize that case in practice.
Several persistent myths have complicated the picture. That VPUs are a universal accelerator, applicable to any workload. That the gains materialize automatically once hardware is installed. That integration is straightforward and the complexity manageable.
None of these hold up under scrutiny.
VPUs excel at standard, high-volume encoding, covering H.264, HEVC, and AV1, where operations are consistent and repetitive. They are not universally optimal. Workloads involving variable bitrates, real-time adaptive streaming, or heavily fragmented content can still perform better on CPUs or GPUs. Deploying VPUs without understanding workload characteristics leads to underutilization, not efficiency. And integration without a native software layer capable of managing VPU sessions, thread allocation, buffering, and recovery at the frame level will leave the hardware chronically underused.
The gains do not come from the hardware. They come from how you use it.
What Scalstrm Is Demonstrating at NAB 2026
At NAB 2026, Scalstrm is participating in the NETINT VPU Ecosystem Pavilion to demonstrate what intelligent VPU integration actually looks like in a production environment.
The centrepiece of that demonstration is Scalstrm’s frame-accurate VPU integration: an architecture that treats VPUs not as external accelerators bolted onto an existing pipeline, but as first-class components of a tightly coordinated, frame-level workflow. Every frame is scheduled, tracked, and routed through a unified control plane that manages VPU sessions, thread allocation, buffering, and real-time recovery. Common bottlenecks such as locked buffers, blocked threads, and mismatched pipeline stages are eliminated by design, not patched over.
The result is higher utilisation, better video consistency, and lower end-to-end latency than loosely integrated deployments consistently deliver.
This is paired with Scalstrm’s just-in-time transcoding model. Always-on pre-encoding is a design choice that no longer makes economic or operational sense. Generating streams only when requested ensures every processing cycle delivers value. Idle compute drops. Storage requirements shrink. Infrastructure costs align directly with actual viewer demand rather than anticipated peak load.
When VPU hardware and on-demand software design work together, the efficiency case becomes measurable and predictable.
Energy Efficiency Is an Architectural Outcome
One thread running through every conversation Scalstrm has had on the road to NAB is sustainability. Rising energy costs and tightening environmental targets are pushing streaming platforms to scrutinise their infrastructure in ways they previously did not need to.
The common misconception is that sustainability comes from buying more efficient hardware. In reality, energy efficiency is a measurable outcome of architecture choices. Platforms that generate streams on demand, orchestrate workloads intelligently across cloud, edge, and on-premises environments, and apply VPU acceleration only where it genuinely adds value consistently outperform those relying on static, always-on pipelines. The savings show up in power consumption, cooling requirements, and the compute hours that simply no longer need to be spent.
Efficiency scales naturally when it is built into the workflow from the start.
A Broader Industry Shift
What the VPU conversation ultimately illustrates is a broader principle: the infrastructure decisions the streaming industry made when scaling up are worth re-examining now that the cost of those decisions is visible. The complexity, the energy overhead, the operational weight of legacy transcoding architectures are not inevitable features of running video at scale. They are consequences of specific design choices, and design choices can be changed.
Scalstrm’s platform exists to demonstrate that directly. Cloud-native, software-defined, built on microservices architecture, and now demonstrably capable of integrating next-generation VPU hardware at the frame level, it is an approach that meets the industry where it is heading, not where it has been.
Meet the Team at NAB
Scalstrm will be present at the NETINT VPU Ecosystem Pavilion throughout the event.
- Patrik Alm, Founder & CEO
- Ola Bengtsson, Founder & CTO
- Dominique Vosters, Business Development and Sales Director
- Pontus Eklöf, Sales Director
Booth W2713. NETINT VPU Ecosystem Pavilion. 18–22 April 2026. Las Vegas Convention Center
Planning to attend NAB Show 2026? You can claim a complimentary pass using code NS8262 and schedule a meeting with the team.
