Video streaming is an industry driven by constant demand for higher quality, lower latency, and cost efficiency. Every decision in a video workflow, whether related to transcoding, storage, or distribution, impacts operational costs and performance. For engineers working in this space, the balance between efficiency and scalability is critical.
At NAB 2025, Scalstrm and NETINT showcased how a just-in-time transcoding model paired with VPU (Video Processing Unit) technology can significantly cut costs, improve energy efficiency, and simplify infrastructure management. This approach is not just an optimization; it represents a shift in how video delivery is conceptualized and executed.
This article explores key takeaways from Scalstrm’s approach, highlighting lessons for engineers looking to optimize their streaming workflows.
The End of CPU-Dominated Transcoding
For years, transcoding has been heavily dependent on CPU-based processing, requiring significant computing power and energy. Scalstrm initially relied on CPU-based VOD transcoding but quickly encountered its limitations. CPU-based transcoding results in high power consumption per stream, an inability to scale efficiently for large workloads, and increased infrastructure costs for on-premises solutions.
To address these issues, Scalstrm transitioned to NETINT’s VPU technology, which offers a hardware-accelerated, energy-efficient alternative to traditional transcoding. VPUs enable higher density by handling more concurrent streams per watt than CPUs or GPUs. They lower cost per stream by reducing both capital expenditure (CapEx) and operational costs (OpEx). Additionally, they optimize power efficiency by using less energy per encoded stream, contributing to improved sustainability.
If CPU-based transcoding is still being used for large-scale applications, there is a significant opportunity to enhance efficiency through VPU technology.
Just-in-Time Transcoding: A Smarter Approach to Storage and Cost Reduction
Traditional video workflows involve storing multiple renditions of a video file, one for every resolution and bitrate needed. This method is inefficient in terms of storage and leads to significant cloud costs. Scalstrm’s just-in-time transcoding model addresses this challenge by dynamically generating renditions as needed. Instead of storing redundant versions of a video, only the highest-quality master file is maintained, while lower-resolution renditions are created on demand.
This approach significantly reduces storage costs, with Scalstrm reporting a 50 to 60 percent reduction in cloud storage expenses. On-the-fly generation of ABR (Adaptive Bitrate) profiles ensures that viewers receive an optimized stream based on their network conditions and device capabilities. Additionally, this method allows for dynamic adaptation to devices and networks without requiring pre-encoded files.
For example, rather than storing ten versions of a video file, Scalstrm’s system only keeps the original 4K master file. When a viewer requests the content, the necessary renditions, such as 1080p, 720p, or 480p, are created as required. This results in substantial savings on storage costs while maintaining flexibility in content delivery.
Voices of Video – on-demand:
Maximizing Efficiency in Video Streaming: Scalstrm’s Just-in-Time Transcoding with NETINT
with Dominique Vosters from Scalstrm
How Scalstrm Cut 40 Servers Down to 2
One of Scalstrm’s customers conducted a real-world case study in which they replaced forty traditional CPU-powered servers with just two NETINT-powered systems. The results demonstrated significant cost savings and efficiency improvements. By making this transition, the customer saved approximately $85,000 per year in energy and infrastructure costs. In addition to the financial benefits, there was an 85 to 95 percent reduction in power consumption, leading to a more sustainable operation. Despite the reduction in hardware, the system maintained the same level of video quality while significantly lowering overhead.
By adopting a VPU-based transcoding model, companies can reduce the number of servers required, cut energy costs, and streamline infrastructure management without compromising performance.
Hybrid Cloud Workflows: Flexibility Meets Scalability
Broadcasters today operate within diverse infrastructures, with some workloads remaining on-premises while others are deployed in public cloud solutions such as Akamai Connected Cloud. This hybrid cloud model provides flexibility and scalability, allowing for cost-effective and efficient content delivery.
Hybrid cloud workflows enable broadcasters to spin up additional resources on demand, particularly for major events such as sports finals, award shows, and live broadcasts. Rather than investing in permanent infrastructure that may be underutilized for most of the year, companies can scale up and down as needed. The model also reduces idle infrastructure costs by allowing non-critical workloads to be offloaded to the cloud while retaining control over essential operations. Additionally, it provides redundancy and failover options, ensuring high availability and reliability for video streaming.
Scalstrm designed its transcoding system to be cloud-native while maintaining the option for on-premises deployment. This approach offers cost control for core workloads while providing elasticity for peak demand scenarios. As more broadcasters move toward hybrid cloud workflows, solutions that support seamless transitions between on-prem and cloud environments will become increasingly valuable.
What We Showcased at NAB 2025
During NAB 2025, Scalstrm demonstrated the power and practicality of just-in-time transcoding using NETINT’s VPU-powered infrastructure. Live demos in the NETINT booth highlighted how this approach reduces server count, slashes energy use, and delivers on-demand renditions without compromising stream quality.
Attendees were particularly interested in:
-
Real-world side-by-side comparisons between CPU, GPU, and VPU-based transcoding.
-
The ability to dynamically generate ABR ladders based on device profiles using Scalstrm’s origin integration.
-
Cloud-native deployments using Akamai Connected Cloud for hybrid cloud workflows.
These conversations confirmed the industry’s growing appetite for scalable, energy-efficient solutions that don’t require compromise on quality or flexibility.
The Future of Codecs: What’s Next After H.264?
Most streaming workflows today rely on H.264 and HEVC (H.265), but AV1 is gaining traction as a next-generation codec. Scalstrm anticipates increased adoption of AV1 within the next one to two years, though several barriers still exist.
Currently, H.264 remains the most common codec due to its widespread support across devices and platforms. HEVC, while more efficient, has licensing concerns that have limited its adoption in certain markets. AV1, on the other hand, is royalty-free and offers better compression efficiency, but its adoption has been slower due to increased computational requirements and limited hardware acceleration.
Scalstrm’s just-in-time transcoding system is codec-agnostic, meaning it can adapt as industry preferences evolve. This ensures that broadcasters are prepared for future transitions while maintaining backward compatibility with existing standards. As AV1 and other emerging codecs become more widely supported, just-in-time transcoding solutions will play a crucial role in enabling efficient adoption.
Embracing the Future of Efficient Video Streaming
The lessons from Scalstrm’s NAB 2025 presence provide a roadmap for optimizing video workflows. Engineers looking to enhance their streaming infrastructure should consider shifting from CPU-based transcoding to VPU technology for improved efficiency. Storage costs can be significantly reduced through just-in-time transcoding, which eliminates unnecessary file storage while maintaining quality. By implementing a VPU-based approach, companies can reduce hardware overhead, lower energy consumption, and simplify infrastructure management. Hybrid cloud workflows provide the flexibility to scale resources dynamically, ensuring cost-effective and reliable content delivery. Finally, as new codecs such as AV1 gain traction, adopting a flexible transcoding system will be key to future-proofing video streaming operations.
For video streaming engineers, efficiency is not just about reducing costs but also about achieving sustainable and scalable performance. Just-in-time transcoding, VPUs, and hybrid cloud architectures are not theoretical improvements but practical, deployable solutions that can transform video delivery at scale.
Schedule a meeting to learn how NETINT VPUs can enhance live streaming with energy-efficient, scalable solutions.
