Short-Termism, The Sustainable Materials Gap, And AI Licensing Agreements For Brands

Renewcell’s Bankruptcy Will Stand As A Cautionary Tale. Will Fashion Heed it Quickly Enough?

This week, Swedish textile recycling pioneer Renewcell filed for bankruptcy after failing to secure sufficient financing to continue its operations. In Renewcell’s home country there are routes to return from bankruptcy, and The Interline is hopeful that the company recovers, but the prevailing sentiment is that this process brings the company at least a step closer to the end of what’s been a winding road.

For context on why this matters, Renewcell has emerged as a lighthouse for industrialising the process of textile-to-textile recycling. Its eponymous process recycled cellulosic textile waste – like worn-out cotton clothes of the right composition, and production scraps that would otherwise be headed to landfill – and transformed it into a new material called CIRCULOSE®. And for an industry where end-of-life garment and textile reuse and the slow roll-out of reduced-impact materials are both hugely pressing challenges, a company that addressed both was bound to attract attention.

So it was no surprise that Renewcell became a major fixture in sustainability conversations – including writing for The Interline’s own Sustainability Report 2023 – or that it attracted backers and buyers including Inditex, Levi Strauss & Co, Ganni, and H&M. Canopy Planet, the NGO advocating for responsible forestry practices, was also a prominent supporter.

On paper, then, this was an ideal picture. A technology / process and a novel material output that a market sorely needed, reaching the point of industrialisation and commercialisation just as the mandate for disclosure and extended-lifespan impact reporting began to feel real for much of Renewcell’s addressable market.

In practice, though, the company entered a “strategic review” period last November, precipitated by a dramatic shortfall in sales versus expectations. And that review culminated in this week’s announcement.

What went wrong? If the maths made as much sense as they seemed to, why did fashion brands and their proximate suppliers not flock to Renewcell’s recycling service and place volume orders for Circulose?

Over at Forbes, Brooke Roberts-Islam (who has also written for The Interline in the past) has an excellent, sympathetic breakdown that questions whether a service like Renewcell’s recycling operations should have been located in consumption markets in the first place, rather than embedded into the supply chain, but that also gets straight to the heart of the matter: more sustainable fibres and materials might perform well, they might align with brands’ and buyers’ values, and they might be expedient to use from an impact assessment perspective, but these do not automatically translate into commercial justification for adopting them at scale – especially not in an industry where the prevailing mindset is a primarily commercial one.

And this is where the industry’s short-termism and its status as a profit-generating machine are laid bare. Here and now, excepting any wider considerations, on a price-per-yard basis so-called “preferred materials” are typically more expensive than the less sustainable alternatives. For commercial enterprises – many of which are publicly traded – this is where the conversation begins and ends; achieving the right margin on a product has remained priority one, with environment and ethics stacked behind it.

But this calculus is destined to change. A 2023 report by Boston Consulting Group, Textile Exchange and Quantis revealed that the “preferred” raw materials demand-and-supply gap could rise to as much as 133 million tons in 2030. This is the net difference between fashion’s need for reduced-impact fibres and fabrics (of which Circulose was an example) and methods of moving the industry towards circularity (of which Renewcell’s recycling business was another example), and the market availability, viability, and scalability of the supply of those things.

Although Renewcell is only one company – its fortunes could be giving us a cautionary preview of how short-term costing is set to clash with mid-term strategy and, eventually, business continuity.

For instance, N Brown Group  (which owns Simply Be, JD Williams and Jacamo brands) this week committed to reduce scope 1, 2 and 3 greenhouse gas emissions from its operations by 46% by 2030, compared with its 2021 level. That target has now been approved by the SBTi, which assesses companies’ targets and ensures they are in line with the Paris Climate Agreement. Other UK retailers that have had their net-zero targets validated by SBTi include John Lewis, THG (previously The Hut Group) and Burberry.

These are clearly not targets that these brands intend to walk back from, but as more organisations look to meet a similarly high bar, the squeeze on the supply of lower-impact fibres, materials, and more environmentally and ethically sustainable production partnerships is going to become acute.

And greater demand plus constrained supply does not exactly have a long history of driving prices down. Which is a glib way of us pointing out – as many other industry commentators and analysts have done – that while the immediate news is related to a single company, it’s also indicative of an industry that is proving slow to adopt and support the innovation that will be necessary to safeguard its future operations.

Within a very short timeframe, there is an “innovation gap” that needs to be bridged if the sector is to meet its 45% greenhouse gas emissions reduction target for the production of raw materials by 2030, in line with the Paris Agreement. And the evidence before us this week is that this gap is likely to be widening, not closing.

AI’s Expanding Reach Keeps Pushing Brand Boundaries

Things move quickly in AI – and not just in the sense of new and better models being introduced. As it becomes clearer that large models will rely on hoovering up multi-modal data from across the internet (after which their training will need to move to synthetic data, which The Interline will analyse in our upcoming AI Report), more content platforms, hubs, and aggregators are choosing to licence their assets to AI companies as a way of being compensated for scraping that seems inevitable.

This week the parent company of Tumblr and WordPress.com,  Automattic, is allegedly set to strike a deal  to sell user-generated content to artificial intelligence companies OpenAI and Midjourney, to help train their AI models.

Automattic is staying tight-lipped, responding to media requests with a published statement claiming: “We will share only public content that’s hosted on WordPress.com and Tumblr from sites that haven’t opted out.” The statement notes that legal regulations don’t currently require AI companies’ web crawlers to abide by users’ opt-out preferences. The ultimate line of Automattic’s statement hints at the reported deals, saying, “we are also working directly with select AI companies as long as their plans align with what our community cares about: attribution, opt-outs, and control.”

AI data training deals have become a lucrative opportunity for websites treading water in today’s slippery online publishing landscape (Tumblr’s staff was reportedly reduced to a skeleton crew in late 2023.)

Last week, Google struck a deal with Reddit (ahead of the latter’s IPO) to train on the platform’s vast knowledge base of user-created content. Meanwhile, OpenAI rolled out a partnership program last year to collect datasets from third parties to help train its AI models. This has been controversial for some users, who feel uncomfortable about their content – on Tumblr, this is often personal writing or photography or art – being used to train AI, but for those users the growing consensus among platform holders is that licensing is the only viable route to monetisation in a near-future era where the need to actively visit content platforms is reduced, with AI models surfacing the relevant results instead.

Beyond the avalanche of user-generated content that is now falling within AI companies’ reach (much of which, to be clear, was already scraped during the training of various large models, making this a deal for ongoing content flow, rather than retroactive licensing) there is also an angle for fashion brands to consider here.

While most brands will have proactively sought to stop AI crawlers from scraping their own websites, they have precious little control over how other platforms approach AI monetisation with content that incorporates their products. Consider curators, influencers, product reviewers, street style blogs, and so on – all of which, if hosted on centralised platforms that permit AI companies to train on their data, will be an indirect route by which brands’ newest products will wind up in large model training datasets anyway.

In light of what feels like an inevitability, could we see brands actively pursuing these kinds of training licenses? It sounds deeply counter-intuitive today, but it could, in theory, mean greater control over how their content is used and shared, ensuring that their work is properly attributed and compensated.

With concerns about the profit motive behind AI development, control, misuse, and inclusivity looming large in the AI sphere right now, brands are likely to be essentially forced to balance their interests with the demands of a new kind of internet – one where they are able, if they choose, to use AI to their advantage, but one where they are set to be forced to make hard choices about how to safeguard their brand integrity. 

The best of The Interline

Kicking off this week, Safir Bellali asks what history can teach us about the right way to approach digital storytelling? And what possibilities the new generation of 3D tools and rendering engines can unlock.

We talk to Centric Software’s Director of Product Management on the enterprise-wide strategic opportunities opened up by digital product creation.

In her first exclusive for The Interline, Sophie Benson explores how fashion keeps critical sustainability indicators close to its chest, creating a lack of clarity around material composition and the scale of (over)production.

Mirroring a wider mingling of top-down technology deployments and grassroots disruptions, Sylwia Szymczyk writes on how success in DPC will only be secured with the buy-in of a broad spectrum of different teams, job roles, and disciplines.