Zapping cuts server footprint 75% using live streaming with NETINT VPUs

The shift to dedicated video processing units (VPUs) demonstrates that independent streaming platforms can achieve broadcast-level latency and 4K quality without the massive capital expenditure typically required for data center expansion. By collapsing the physical and electrical footprint of the encoding ladder, Zapping has redirected resources toward its proprietary CDN and regional expansion. This move highlights a growing industry trend where specialized silicon replaces general-purpose compute to manage the high-bitrate demands of live sports. As the platform scales into four additional countries, watch for Zapping to transition its HEVC workflows toward AV1 to further reduce bandwidth costs by an estimated 40%.

NETINT Technologies has been expanding its VPU footprint across live streaming operators that need to compress more channels per rack unit without proportional increases in power or cooling. In June 2026, Ericsson launched its AI in RAN commercial software subscription claiming up to 20% higher downlink throughput across more than 15 live deployments, illustrating how the broader telecom and streaming infrastructure market is shifting toward specialized silicon and software-defined acceleration rather than general-purpose compute. NETINT's Quadra T1U VPU, which Zapping deployed for its HEVC encoding ladder, competes directly with GPU-based transcoding from NVIDIA and FPGA solutions from Xilinx (now AMD), but differentiates on power-per-stream metrics that matter most at the edge of content delivery networks serving latency-sensitive sports audiences.

The business case for dedicated video silicon is being reinforced by cloud provider economics. Nokia announced a partnership with AWS and Databricks to build a unified data and control layer for autonomous network operations, signaling that hyperscale cloud platforms are actively courting telecom and media workloads with integrated orchestration stacks. For OTT platforms like Zapping that operate proprietary CDNs across multiple Latin American markets, the decision to run NETINT VPUs on-premises rather than renting GPU instances from AWS or Azure reflects a broader pattern where high-volume, always-on encoding workloads favor capital expenditure over operational expenditure once channel counts exceed a threshold. Ericsson's own agentic AI blueprint, which runs on Amazon Bedrock and positions more than 20 cloud-native AI applications across OSS and BSS functions, shows that even vendors deeply embedded in AWS ecosystems acknowledge the tension between cloud flexibility and the cost predictability that dedicated hardware provides for sustained workloads.

On the technical side, NETINT's T408 and Quadra T1U VPUs are positioned as drop-in replacements for GPU-based encoding in scenarios where stream density and power efficiency dominate the decision matrix. The competitive landscape for hardware-accelerated video processing has intensified as Nokia and Ericsson diverge sharply on their AI-RAN strategies, with Nokia building its entire Layer 1 RAN on NVIDIA CUDA and GPUs while Ericsson optimizes existing baseband silicon. That same philosophical split, general-purpose GPU versus purpose-built accelerator, is playing out in the video encoding market. Zapping's reported 90% reduction in encoding power consumption aligns with independent benchmarks that have shown ASIC and VPU-based encoders consuming between 5 and 15 watts per 1080p60 stream compared to 150 to 300 watts for equivalent GPU-based pipelines, a gap that compounds significantly at the scale of a multi-market live sports platform.