Harmonic, NVIDIA, and Akamai target sub-second live sports latency
Harmonic is collaborating with NVIDIA and Akamai to enhance real-time video delivery, particularly for live sports and events. The partnership leverages NVIDIA's NVENC/NVDEC for AI-driven video quality improvements and Akamai's Inference Cloud with NVIDIA RTX 6000 Pro Blackwell GPUs for low-latency edge compute. This aims to reduce latency and improve video quality across multiple codecs using AI models at the edge.
Key Takeaways
- Uses NVIDIA RTX 6000 Pro Blackwell GPUs on Akamai’s global network to decentralize video inference tasks.
- Integrated EyeQ technology employs trained CNN models for real-time rate control and encoding mode decisions.
- Supports multiple professional codecs including AVC, HEVC, AV1, and MV-HEVC for spatial video applications.
- Edge-based execution enables local AI functions like speech-to-text and sports highlight detection for live environments.
Why It Matters
The shift from post-production AI to real-time edge inference is critical as live sports increasingly move to OTT platforms. By executing complex models on Blackwell GPUs close to users, Harmonic avoids the 45-second 'spoiler gap' caused by centralized cloud processing. This architecture effectively bridges the gap between broadcast stability and interactive streaming features. For the ecosystem, it signals a move away from generic cloud compute toward specialized, high-memory GPU instances optimized for the 'inference era.' Watch for benchmarks comparing GPU-accelerated edge encoding against traditional software-only stacks for latency and bitrate savings.
Additional Context
The push for edge-based inference comes as viewer expectations for live sports hit an inflection point in 2026. Per Streaming Media Connect (March 2026), reducing glass-to-glass latency remains a top business priority for rights holders like DAZN and BT Group to prevent social media 'spoilers' from ruining the viewer experience. Market data from StreamingMetrics in April 2026 indicates that 70% of viewers would consider switching providers for a lower-latency stream, with a growing preference for delays under 50 milliseconds for synchronized betting and interactive features. Akamai’s pivot toward AI infrastructure is underscored by its announcement in March 2026 that it has acquired thousands of NVIDIA Blackwell GPUs. This deployment powers the Akamai Inference Cloud across more than 4,400 global locations. According to Akamai (May 2026), this distributed architecture can reduce AI inference costs by up to 86% compared to traditional hyperscale clouds by minimizing data egress and centralizing compute only for model training, while moving execution to the edge. Harmonic’s use of EyeQ technology represents a broader industry trend toward 'intelligent' compression. As revealed at the 2026 NAB Show, Harmonic is now orchestrating third-party AI functions directly within its video appliances. This allows for automated sports highlight creation and scene analysis to be performed during the live stream. In early 2026, benchmarks of the NVIDIA RTX 6000 Pro Blackwell edition showed its 96GB of GDDR7 memory is specifically optimized for these high-throughput media pipelines, enabling simultaneous 4K frame-rate doubling and AI noise reduction without increasing CPU load.
Read full article at harmonicinc.com