How Zapping scaled premium low latency streaming using NETINT VPUs across Latin America
Customer Case Study | IBC 2026
AT A GLANCE
Zapping built one of Latin America’s fastest-growing live OTT television platforms by optimizing video infrastructure for density, low latency, and operational efficiency. As subscriber growth accelerated, the company needed to reduce rack space, power consumption, and encoding costs without compromising quality.
This customer case study explains how Zapping scaled Live Streaming with NETINT VPUs, reducing its server footprint from approximately 100 CPU-based servers to 25 VPU servers while lowering encoding power consumption by about 90% and supporting continued expansion across Latin America.
Figure 1. Zapping’s path on NETINT, from first VPU to roadmap.
Figure 2. Zapping’s path on NETINT, from first VPU to roadmap.
The Company
Figure 3. Ignacio “Nacho” Opazo, Co-founder & CTO, Zapping
Zapping set out to do what incumbents said could not be done profitably: deliver premium, cable quality live television, including the region’s most watched soccer, over the internet, with lower latency than local broadcast channels. Founded in Chile and led by co-founder and CTO/CPO Ignacio “Nacho” Opazo, Zapping positions itself as much a technology company as a content company. It built its own encoders, low latency delivery architecture, custom apps, and proprietary CDN. As Nacho puts it:
Zapping is the Netflix of live streaming here in Chile, in Latin America. We developed all our technology, the encoders, our low latency, and the apps on each platform. We developed our own CDN; I think it is bigger than Akamai and Fastly here in Chile and Ecuador. We are taking the same steps as Netflix: you make your platform, you make the UI, you make the encoding process, and then you deliver.
That vertically integrated approach turned Zapping into one of Latin America’s fastest growing live television streaming platforms and, in the process, made encoding density, power, and cost a competitive advantage.
The challenge: soccer, latency, and the economics of scale
In Latin America, live sports, soccer above all, drive subscriptions. They also create demanding engineering problems, starting with latency. “Here in Chile, the soccer matches are premium. You can hear your neighbor screaming if they have a cable operator with lower latency,” Nacho explains. “In Brazil it is more complicated, because some matches are free to air. So we must be fast enough to try and match free to air.” Soccer is punishing to compress, with wide shots of the field where players are small but jersey numbers must stay visible. With crowds, waving flags, and fast pans, preserving detail and color integrity without inflating bitrate is essential.
We developed our own upscaling technology to improve sharpness and achieve superior quality compared to what broadcasters offer. We can upscale images to 4K and convert SDR to HDR.
“Video files are huge, and when you have a soccer match, thousands of people hit your servers, saturating the region’s internet.” In some markets, inter ISP trunks were as small as 10 Gbps, and Zapping would overload them. That prompted the company to build its own CDN and implement a server cache system within every possible ISP across Latin America. Every one of those requirements had to be met affordably enough to install and operate data center capacity across South America, market by market. This made channel density per rack unit, power draw, and heat dissipation high priority problems to solve.
The journey: from CPU to GPU to VPU
Nacho tested every computing architecture, CPU and GPU, before settling on VPUs. What he ruled out is as instructive as what he chose.
Figure 4. Density per 1RU decided it: the VPU delivered what CPU and GPU could not.
CPU (Intel Quick Sync) delivered good quality, but “my problem was getting more density for the rack unit. Intel enabled just five streams per 1RU. Plus, the power and heat produced was really high.”
GPU (NVIDIA P2000 to T4) was an improvement, with roughly 50 complete encoding ladders per 1RU on a dual T4, 80 core configuration, “but still insufficient.”
VPU (NETINT Logan to Quadra series). Nacho discovered a NETINT article showing he could achieve 122 channels per rack unit using the first generation T408. “I found the power draw was really low, as was the latency, and the quality of both H.264 and HEVC is really good.” Integration, a common worry with dedicated silicon, was a non issue: “We control the encoders with FFmpeg, and converting over to the NETINT transcoders was really seamless. Really, really easy.”
Zapping moved into production on T408 based streaming servers in 2022 and 2023, then transitioned to the second generation Quadra T1U VPUs in 2025. This provided the density and codec headroom needed to keep pace with expansion into larger markets like Brazil. As Nacho framed it: “I need to get more dense. Brazil is a really big country. We need more power and more density in the rack.”
The density math is what changed the business. Using NETINT VPUs, Zapping reduced the server count needed to serve its ladders from roughly 100 CPU based servers to a fraction of that.
We currently have 25 servers of varying density depending on their purpose, using T408 and Quadra VPUs. Each server is serving 50 to 60 channels with full ABR ladders consisting of 6 to 8 profiles using H.264 and HEVC.
The results
This new configuration allows us to reduce the number of rack units, energy costs, and capex, while giving us high availability and total control over the quality of our encoding, surpassing our competition in image quality and customer satisfaction.
Density and footprint. Zapping went from about 100 software based servers to 25 VPU servers for the same encoding volume, roughly 75% fewer servers. Power and sustainability. Previous infrastructure consumed about 600 watts per server, while the VPU based servers require just 250 watts and provide three to four times the channel density. Across the fleet, that works out to roughly 90% less power for the same encoding workload, which directly reduces both cost and Zapping’s footprint in every data center it deploys.
Figure 4. The efficiency gain: same workload on a quarter of the servers..
Cost. Reducing the server count from more than 100 to 25 for the same volume of encoding, together with the drop in power, drives substantial capex and opex savings. As Nacho puts it, NETINT has helped “accelerate our expansion while increasing our profitability.” Latency. By placing a NETINT server inside each broadcaster’s CDN, Zapping originates its streams as close to the source signal as possible. In practice, Zapping lands within a few seconds of free to air, and in some cases ahead of it.
Figure 5. Encoding inside each broadcaster’s CDN keeps Zapping close to the source signal.
Bandwidth via HEVC and AV1. “HEVC is really important. Using these advanced codecs, we achieve a 40% lower bitrate than H.264 at the same quality; the user can see the difference.” Alongside Super Resolution (4K upscaling), Zapping is now beginning to work with AV1 and hopes to replace HEVC in the near future.
During the 2026 World Cup, our CDN reached consolidated traffic over 2 Tbps.
Why it matters for streaming video platforms
Zapping is proof of a thesis at the heart of the VPU Ecosystem: that purpose built video silicon lets an independent operator compete with, and outperform, far larger incumbents. By collapsing the cost, power, and rack space of encoding, VPUs let Zapping put capital into content and its own CDN instead of ever larger data centers.
NETINT servers are an affordable, functional, and high performance element of our success: unparalleled density, excellent low latency, and H.264 and HEVC quality with extremely low power consumption. NETINT has helped accelerate our expansion while increasing our profitability.
Zapping continues to advance its software that clusters VPUs and distributes video encoding jobs across its data centers, while unifying these flows with its Stream Gateway. Its Upscaling and SDR to HDR (Super Resolution) technology demands ever more efficient encoding, so the team is now optimizing encoding using neural networks
With VPUs, we continue to open new countries in Latin America, and hope to open four new countries in the medium term.
