[VIDEO] Glacier and the evolution of AI and data-driven technology in recycling

Is a new era, where environmental goals meet the precision and efficiency of advanced manufacturing, emerging in recycling? Material recovery facilities (MRFs) across the U.S. are beginning to adopt AI-driven systems that offer real-time operational insights, transforming recycling from a labour-intensive, inconsistent process into a streamlined, data-rich operation. 

Among the innovators shaping this transition is Glacier, a San Francisco-based startup leveraging affordable AI robotics to help recyclers improve recovery rates, enhance bale purity, and unlock new value from the waste stream.

"Recycling is manufacturing," says Rebecca Hu-Thrams, Glacier co-founder and CEO. "These MRFs are responsible for taking in raw feedstocks, producing a certain yield, producing at a certain quality bar. At the end of the day, you're making new stuff out of raw materials."

From feel-good recycling to circular manufacturing

When Glacier was founded in 2019, the recycling industry was facing mounting pressures. Changes like China's National Sword policy, evolving packaging materials, and rising operational costs were making traditional recycling practices increasingly untenable.

"We saw an opportunity to apply some of these really incredible cutting-edge technologies like AI, like purpose-built robotics, to actually come in and provide some of that technology to act as a bridge to help MRFs meet these mounting demands and actually turn those into opportunities," says Hu-Thrams.

Glacier's approach uses AI not just to sort recyclables more efficiently, but to reframe how facilities view their role: as manufacturers operating complex, data-driven production lines. Their AI vision systems offer real-time reporting on the composition and quality of material streams, allowing MRFs to track recovery rates and contamination with enhanced accuracy.

"One of the incredible things about this type of AI technology is its ability to continually learn and improve," explains Hu-Thrams. "Unlike other types of legacy sorting equipment, where once you install it, that's kind of the job that it does until it becomes obsolete, we can actually teach our AI model to recognize [new packaging] very quickly. Suddenly, the robot is picking stuff that wasn't even around when it was installed."

Making smart recycling affordable and scalable

A major barrier to automation in recycling has traditionally been cost and complexity. Glacier set out to remove those obstacles by striving to design a solution that is faster, cheaper, and easier to implement than conventional systems. According to Glacier, their system costs less than standard options, requires just three feet of conveyor space, and can be installed in one day or less, avoiding facility downtime. 

"A typical install is extremely fast, all things considered. It usually takes less than a day," says Hu-Thrams. "We usually go in on the weekend when the MRF's not running, so there's literally zero downtime for that installation."

Cost is another differentiator. Hu-Thrams says Glacier's AI recycling robots cost approximately half as much as traditional automation options. The company also structures contracts to incentivize buy-in. 

"We'll put most of our fees at risk," Hu-Thrams says. "The vast majority of what we're owed, you don't actually pay until we can demonstrate to you that we're hitting a certain success metric that the customer and Glacier have collectively defined."

The scalability of the system extends beyond installation. "We've already [increased our manufacturing capacity eight-fold] just to keep up with the demand that's coming in the door," says Hu-Thrams. She also notes that this allows Glacier to invest even more heavily into delivering robots faster, providing even more support and coverage for MRF operators.

Real-time AI insights unlocking new value streams

Beyond improving bale quality, Glacier's technology is helping MRFs recover valuable materials that often slip through with traditional processes. Their vision systems have uncovered significant opportunities on residue lines — the material typically destined for landfills.

"We've seen that in recent years there's been increasing rates of contamination from things like plastic film [and] flexibles or other materials that make it hard to keep paper bale quality high," says Hu-Thrams. "And that residue line is a gold mine of opportunity."

Previously, some facilities resorted to labour-intensive methods like manually reviewing GoPro footage frame by frame to tally missed recyclables. Glacier's AI vision system automates that process. 

"Being able to put Glacier's vision system in there and report out [in real time] on every single item that's on the residue line or that's on a QC line or a container line really helps them to automate the process of better understanding material flow," says Hu-Thrams.

Just as important as the data itself is how it's delivered. "The feedback our customers give us all the time is they're often on their feet dealing with literal fires in their plant, and they don't have the time to sit down and try to understand a thousand-row spreadsheet," Hu-Thrams explains. "What they need is really actionable insights."

The future of smarter, more profitable recycling

Glacier's broader ambition is not limited to robotics. "Glacier's mission from the start was never to be a robotics company or an AI company," says Hu-Thrams. "It's really to build what we call circular manufacturing." 

The company envisions expanding its AI and vision technologies into adjacent waste streams, including electronic waste, organics, and even construction and demolition materials.

"One of the really wonderful advantages about having built a sort of purpose-built technology stack for recycling is that we've designed our technology with the anticipation of being able to scale it into numerous adjacent waste streams in the future," says Hu-Thrams.

The company is also investing in expanding its AI capabilities. "We're investing in making our AI model capable of detecting increasingly more and more different types of commodities," says Hu-Thrams. "We're already able to detect about 90 percent of the curbside recycling stream, so we're tackling that long tail of 10 percent."

To support this growth, Glacier recently closed a $16 million Series A funding round led by Ecosystem Integrity Fund, with participation from Amazon's Climate Pledge Fund and other investors. The funding will fuel expanded deployments and accelerated product development.

"This year, we're also investing heavily in really bringing the possibility of those actionable data insights to market," says Hu-Thrams. "Now that we have this massive amount of data, we've got over three billion proprietary waste images, how do you actually help your end users make heads or tails of that so that they're not stuck staring at a spreadsheet, but can actually go and immediately make changes that yield better operations and more revenue?"

Looking ahead, Glacier is also exploring how its data could eventually support broader circular economy initiatives, informing producers and policymakers alike.

"Glacier's ultimate goal is that the data can be a helpful source of truth for anyone who's involved in the circular economy," says Hu-Thrams. "We really need inputs from our MRF customers, from our producer customers, and from legislators telling us what they actually care to understand about the recycling stream."

Ultimately, Glacier defines its success by the success of its customers.

"At this juncture, success for Glacier means helping our customers run more efficient, more profitable businesses," says Hu-Thrams. "We find that the best way to make sure that we're serving our customers' needs is to spend time out in the field, actually walking the lines with our MRF customers, and talking to them to understand what it is [that] they're solving for at any given moment."

As the recycling industry continues to evolve, companies like Glacier are showing that the future of recycling isn't just about diverting waste. It's about building smarter, more resilient systems that can find and leverage the value in what was once considered trash.