Pixop Joins Greening of Streaming AI Study

Pixop integrates AI video tech into Greening of Streaming's OWL platform to measure sustainability gains from AI-driven UHD upscaling.

14 Jul 2026

Video is heavy. A single hour of ultra-high-definition streaming can demand far more bandwidth and processing than most viewers ever consider. Multiplied across millions of households, that adds up to a real infrastructure bill, paid partly in electricity. Artificial intelligence, by sharpening video after the fact rather than shooting and shipping it in native UHD, has long been assumed to lighten that bill. Nobody had actually measured it.

Now that is changing. Pixop, a video-enhancement firm, has joined Greening of Streaming, an industry sustainability initiative, in a partnership announced on July 10th 2026. Feeding Pixop's machine-learning upscaling technology into Online WattLab, a platform built to measure streaming's energy use in the real world rather than in theory, the two will compare three approaches directly: AI-driven upscaling, device-side processing, and native UHD production. Whichever wins loses nothing to guesswork.

The stakes are not trivial. "The industry has largely been working from assumptions rather than real-world data on device-side energy avoidability," says Ben Schwarz, president of Greening of Streaming. For Morten Kolle Christensen, Pixop's chief executive, the tie-up is a natural extension of the firm's existing mission. "Delivering best video quality at lowest infrastructure cost has always been central to our mission," he says. "This research adds a meaningful environmental dimension to that equation."

Should the data show AI upscaling to be the leaner option, streaming platforms gain a business case that regulators, not just marketing departments, might actually credit. Europe's broadcasters, already under pressure to show measurable progress on emissions, would have a concrete lever to pull. If it does not, at least the industry will know, rather than assume, where its energy actually goes. Dull as that distinction sounds, it may matter more than any single upscaling algorithm.