The State of Streaming Sustainability 2026
Welcome to another year of streaming growth, which means another year of balancing growth demands with necessary and innovative sustainability requirements. In this article, I’m going to revisit several of the key themes from The State of Streaming Sustainability 2025 and also explore a few key catchphrases and terms (AI, anyone?) that have dominated the sustainability conversation in the past few months.
Is Sustainability Going Out ofFashion?
During a recent discussion with Barbara Lange, former SMPTE executive director and owner of sustainability consultancy Kibo121, we talked about the perception that 2025 might have been a year of “treading water” on sustainability efforts. A few days later, Lange posted an interview she’d done with Dalet at IBC 2025 on that very topic, noting that 2026 will be a strategic one for sustainability in streaming. “Some think #sustainability is no longer relevant in these chaotic times,” writes Lange in her post. “I believe that it is the exact time to look for efficiencies that not only save energy but also serve your sustainability goals.”
Kibo121’s Barbara Lange talks streaming sustainability with Dalet at IBC 2025.
Lange’s interview highlights the holistic nature of sustainability efforts as they relate to business needs. “The more we can make sustainability a natural part of business strategy,” she says, “the easier it becomes.” Lange goes on to make a seminal point: Even though streaming isn’t the biggest contributor when it comes to overall emissions compared to other, more recent entrants, such as the heavy emphasis on machine learning (aka AI), she notes, “We do contribute and, therefore, we need to do our part” in solving the problem. Lange also frames a part of the solution in terms of energy efficiency, an area near and dear to the heart of the Help Me Stream Research Foundation, of which I’m the founding executive director. “When you talk about sustainability, people’s eyes sort of glaze over. But when you talk about energy efficiency, [sustainability] is really about efficiency gains,” she says.
Speaking of AI …
If there’s any one term that’s absolutely permeated the streaming industry—and, frankly, almost every other compute-intensive industry—over the last year, it’s AI. Having been around the industry for the prior three waves of AI buzz, this fourth wave seems to have both the highest hype factor and the highest potential. In many ways, there are a number of very practical use cases for fourth-generation AI solutions, building on several innovations in prior waves—such as speech-to-text and recommendation engines—to yield practical benefits like real-time translation into multiple languages.
The defining difference in this wave is the sheer size of the learning model libraries, brought about by an overall reliance on cloud-based storage of practically everything from emails to position papers to gobs and gobs of media. The streaming industry is no exception, as large libraries of content—original, mezzanine, and user-generated, as well as live archives—are indexed against a number of weighted averages around popularity, best-practice references, etc.
To cut through the hype, Greening of Streaming, which was founded by Dom Robinson and is now run by president Benjamin Schwarz and a volunteer secretariat, released a position paper in January 2026 titled Artificial Intelligence in Streaming Media Sustainability: Distinguishing Impact From Innovation. “Our position is straightforward,” it notes, after giving a detailed context to the terminology around AI, including convolutional neural networks (CNNs). “AI in streaming workflows is neither inherently a sustainability solution nor an inherent threat. Its net impact depends on where it appears in the workflow, what type of AI is deployed, and how much energy it adds versus avoids.”
While this might, at first glance, seem like a really long caveat, the paper notes that Greening of Streaming members “explicitly distinguish [streaming AI] from large language models (LLMs), which, while energy-intensive, have limited direct application in traditional streaming workflows.”
Greening of Streaming is also careful to note what exactly the organization measures. “In this paper, sustainability refers primarily to energy consumption and related carbon impacts. Broader sustainability dimensions—such as water use, hardware lifecycle, and social impacts—are acknowledged but fall outside our current measurement scope.” The limited scope, according to Greening of Streaming, also “avoids duplicating work already underway in other organizations (SVTA, CTA, and data center sustainability groups).” Then the paper offers a good synopsis of three current areas where AI-based solutions are marketed: encoding and compression, CDN and delivery optimization, and content and quality management.
On the encoding front, content/context-aware encoding (CAE) has been around for a decade, and it’s not inherently an AI-driven process. However, the use of the term AI for CAE is probably adequate for the majority of our industry, somewhat akin to saying “then magic happens” in describing a technical process to an elderly relative.
The same is also true for per-title and per-shot encoding, but here, the heavy lifting is enhanced by machine learning and weighted-averaging algorithms across a much wider spectrum of content categories. The learnings have come from best practices in a variety of categories and genres (outdoor team sports in day-light versus under artificial lighting, as one example). What is new, however, is a crop of “AI-based compression” codecs that merge together per-shot and CAE encoding with complementary decoders. The industry has been chasing the ability to achieve greater compression while not taxing the consumer’s mobile de
vice with inordinate hardware and battery usage. On the delivery front, predictive caching and intelligent traffic routing can minimize latency and server load for live and on-demand content, respectively. When it comes to content and quality management, not only can machine learning workflows help monitor predictors of real-time QoE, but they can also aid in a more automated approach to compliance, whether for ad delivery or QoS. They can even help to automate the generation of service-level agreements between the content owner and the network service provider.
Too Much Data?
Traffic shaping can also benefit from fine-tuned weighted-averaging algorithms, but one area where this might unintentionally cause problems is an over-abundance of data. Sarge Sargent, a longtime streaming media analytics consultant and co-founder of Sargeway, LLC, highlighted one of the cautionary tales in using AI from a streaming analytics standpoint in an interview at Streaming Media 2025. “One of the problems with traditional dashboard construction is that they may be looking at the wrong data or at the wrong period of time for the data,” he said.
Sargent’s solution? Using agentic AI to build bots that “will consistently mine this data for you so that at any moment in time, you can pull up a view—that’s generated on-the-fly—of that dataset so you can make some decisions on it.”
Streaming veteran Sarge Sargent discusses the downside of traditional dashboards at Streaming Media 2025.
In a follow-up interview in early 2026, Sargent elaborated on the reasons to narrow down analytics to a digestible level. “For several years, the organizations I’ve worked with have leveraged AI for analytics,” said Sargent, who has had stints with Adobe, Disney, and others. “But one of the unintended consequences of measuring almost everything there is to measure is that the dashboards themselves can become overwhelming.” He went on to add that “a novel use of AI these days would be to generate a dashboard of dashboards as a way to visualize at a high level, but also to be able to drill down into minute measurable details should an anomaly arise.”
Power to the People
This Pareto chart strategy and the anomaly detection workflow are valid ways to approach key performance indicators, measuring everything but reporting just the KPIs themselves in a quest for more sustainable streaming. The power requirements that come from a combination of streaming and machine learning, however, seem excessive—so much so that just the idea of planning a data center in certain parts of the U.S. has yielded both organized and grassroots resistance to the idea. It’s not just pushback on AI data centers or cryptocurrency “farms,” but even an aversion to general data center buildouts.
One factor in this pushback is the sheer amount of noise that is generated by poorly designed data centers. It’s become such an issue that legislators, on both the national and local levels, are implementing strict noise-control litmus tests before data center permits are approved.
“Interested developers would … need to meet certain criteria including being 500 feet from a residential district and remaining below certain sound levels,” writes environmental reporter Jorgelina Manna-Rea in an article about potential data centers and crypto mining in Kingsport, Tenn. “Sound and vibration studies would also need to be conducted for the facility. Three sound studies would be required: a preliminary study, an interim study when the facility applies for a building permit approval, and a study six months into its operations." Another factor is potential energy consumption.
The Greening of Streaming position paper references power consumption estimates from the International Energy Agency (IEA), a Paris-based organization focused on power requirements for a variety of industries. IEA notes that data centers have the potential to be one of the largest power use cases on a global basis. However, AI-centric electrical power requirements (referred to by IEA as hyperscalers) will grow at an even faster clip collectively but at a much higher rate when it comes to power consumed by servers, in comparison to overall power consumption in data centers.
Protesters in Lowell, Mich., rally against a proposed Microsoft data center in January 2026.
“China and the United States are the most significant regions for data centre electricity consumption growth, accounting for nearly 80% of global growth to 2030,” an IEA report states. “Consumption increases by around 240 TWh (up 130%) in the United States, compared to the 2024 level. In China it increases by around 175 TWh (up 170%). In Europe it grows by more than 45 TWh (up 70%). Japan increases by around 15 TWh (up 80%).”
Uncertainty also abounds in terms of the cost of power generation. While everyone agrees that there will be a need for more power generation as data centers are added across the U.S., public opinion is torn between the consumer cost of electricity being too high if a data center lands in a particular location—based on the cost to add more power plants, even if they’re nuclear-powered—versus the consumer cost being too high if the data centers themselves cease operation before the energy provider has amortized the power-generation equipment required to run the data centers in the first place.
Whence the Bubbles?
In “The State of Streaming Sustainability 2025,” I referenced a number of factors around data centers, including physical size, cooling, and water challenges. I also noted that there are key geographical “environmental bubbles” in the U.S. that are a good fit for data center operations, even if they may not fit the normal models—massive data centers filling up dormant farmland—that we’ve come to think of as a typical data center.
Those locations still exist and are even more relevant today, as they are well-suited for smaller, more robust deployments. This is in keeping with the interest in smaller—and even micro—data centers noted in past State of Streaming surveys that Help Me Stream Research Foundation has run for Streaming Media. Unfortunately, though, the perceived mismanagement of “big box” data centers has put a crimp into the ability for even these smaller data centers to come online for the benefit of previously marginalized communities.
IEA points out that cooling is a major issue when it comes to data centers overall, albeit not as much of one for AI-based data centers. “The share of cooling systems in total data centre consumption varies from about 7% for efficient hyperscale data centres to over 30% for less-efficient enterprise data centres,” it notes. However, the relevance of cooling-to-overall-power-consumption ratios for hyperscalers may also benefit from higher voltage inflows in addition to the move from AC power in data centers to much more efficient direct DC power.
One approach that’s gaining traction alongside the use of DC power (which itself is estimated to save between 6% and 8% of power consumption due to the elimination of the DC-AD-DC process that’s standard in data center power infrastructures) is the idea of using existing environmental factors to lower overall ambient temperatures. This can be done through employing liquid-cooling solutions in areas that abound in reusable indirect water cooling (as is the case in the majority of nuclear power plants) but also in the use of low-impact or subterranean data center locations. Cryptocurrency mining is probably—pardon the pun—the canary in the coal mine that our industry should be watching. Whether it’s data centers in abandoned subsurface mines, reuse of underutilized forestry facilities (sawmills and paper mills), or oversized Rust Belt power-generation plants that are no longer being used to purpose but are structurally sound and somewhat environmentally shielded, there are a number of ways to set up streaming-centric data centers with minimal new visual or power-generation impact.
What Can We Do?
The question of what we can do permeates many of the discussions at Help Me Stream Research Foundation, as we look for ways to reduce energy consumption and increase compute power on “outdated” servers, tablets, and personal computers in working to fulfill our mission of giving the “gift of streaming” to the more than 1.5 billion people worldwide who have never had internet connectivity, let alone streaming. Whether it’s a potential student intern exploring ways to leverage their STEAM educational training in our applied-learning research labs or a donor looking to fund student internship scholarships and monthly lab costs, the common theme we hear is that—regardless of socioeconomic status, belief or value system, or professional role—the vast majority want to find ways to both prolong the life of “outdated” tech gear and lower overall power requirements surrounding their digital lifestyles.
A well-known European organization, TNO, recently hosted The Energy and Wellbeing Impact of Streaming, a webinar that addressed the practical effects of energy consumption head on. It presented the results of “two research tracks on streaming behaviour and its energy impact, and how these in-sights can be translated into realistic opportunities for change.” The webinar centered on three major areas: self-awareness of power consumption, the practical implications of behavioral change, and overall platform design changes that directly impact energy efficiency. “Streaming behaviours drive energy use,” TNO notes in the webinar description. It adds that the webinar planned to highlight “where users are willing and able to change, and which groups show the greatest potential for impact” and to present several scenarios “based on observed streaming behaviour [to] show where the largest and most realistic reductions can be achieved.”
Into the Flow
Sustainability in streaming is a topic we’ll follow closely in 2026, because the streaming industry is a large consumer of power and not always the best example of energy efficiency when we’re in a growth cycle. Having said that, though, and harkening back to the Greening of Streaming position paper, it’s worth noting that the industry consensus is that streaming itself may lower its overall power consumption if highly targeted AI-enhanced workflows are properly implemented. This provides a flicker of hope that in 2026, we’ll move beyond “treading water” and into the flow of sustainable production and delivery.
Related Articles
In this article, I'm going to look at a few of the underlying issues facing the industry when it comes to sustainability, starting with one of the biggest power consumption challenges of our generation: the explosion in data center growth.
28 Mar 2025
How do others see us, and how do we measure our progress? New data on streaming power consumption will allow the industry to target further reductions in power consumption while working toward longer-term solutions that reuse old technologies alongside current best practices as a way to extend the life of streaming tech for years to come.
18 Mar 2024
The time has long since passed to turn our attention to making the streaming delivery ecosystem more environmentally friendly. Here are some of the key challenges, as well as some suggested first steps towards solutions.
26 May 2020