Cooperative Interventions in the AI Stack. A Commentary on Infrastructure and Power by Phakin Nimmannorrawong
Preparatory Comments by Phakin in the context of the event Cooperative Digital Infrastructure & Data Centers
April 24, 10:00 AM ET – 11:30 AM ET
This essay offers commentary on Tara Merk’s and Kenzo Soares Seto’s research in relation to my work on Bread Cooperative. I should preface this by saying that I myself am not an AI expert, and my research interests lean more toward the sociology of digital communities and collective actions among technologists. Accordingly, I will focus on what both works have taught me, and what I reckon as valuable and instructive for those of us who believe in cooperativism, to better grapple with the question of AI.
Let me begin with a little anecdote here. Whilst preparing for this talk, I was kindly advised not to come in with “smart” questions, and to focus instead on something audiences can actually learn from. However, I’ve realized that we cannot arrive at grounded, useful lessons without asking questions first. Just, certainly, not the “smart” kind.
Tara and Kenzo, along with other essays by ICDE fellows I will cite here, have taught me what are the questions the cooperative community should be carefully considering regarding our relationship with AI.
We know that AI is here to stay, and everywhere it is in extractive (supply) chains. We also know that the current consolidation of AI power within a handful of tech companies is arguably detrimental to cooperatives employing AI at both technical and operational levels. Essentially, extractive AI is built upon an extractive business logic where only what is quantifiable matters, profit is subordinate to nothing, and where profit comes from — and at what cost — is determined by its shareholders alone. AI in its current form stands against cooperative values such as care and accountability, democratic governance, and concern for community.
But rather than going back and forth over whether cooperatives should use AI, Tara and Kenzo would ask: how can cooperative practitioners engage with (build, own, govern, utilize, foster) different layers of AI stack in such a way that stay true to our cooperative values? And what support do we need so that cooperative alternatives can emerge and remain true to their values over time?
By tracing the history of GAD eG, Tara shows that cooperative ownership of data centers, which is constitutive of AI infrastructure, is economically viable, particularly with partnership and support by public institutions. This model provides an alternative framework for thinking about and practicing Solidarity AI, and perhaps AI Sovereignty. If the ownership, and thus the governance structure, of AI data centers determines whose resources are used, how they are used, and what counts as a desirable outcome, then the answers to these questions would shift significantly if such powerful, energy-hungry infrastructure come under the ownership of local people in the communities where it is physically grounded.
As another ICDE fellow recently pointed out, AI infrastructure is technical, institutional and political all at once. Reconfiguring AI’s ownership then means rearranging who “decides when it decides” and what carries more weight than its conventional playbook of “efficiency, optimization, scale”.
Similarly, Kenzo highlights that true ownership of AI cannot be achieved without the control of its infrastructure. Using a vibrant cooperative ecosystem in Brazil as a case study, he emphasizes a critical role of public institutions and policies in providing support to cooperative initiatives at the legal and regulatory level. From technology transfer and knowledge support, to preferential public procurement and fiscal and financial incentives, these regulatory measures, once carefully implemented, can become one of the fundamental scaffoldings that enables cooperative alternatives to take root and stand firm against profit imperatives and the encroachment of tech giants creeping across the Global South.
To protect our cooperative values, both projects provide a compelling case in treating AI not as a one unified thing, but as the interconnected layers of technology that are always deeply embedded in specific labor systems and the material resources of our world. They also highlight the gaps within each layer of the AI stack where the cooperative community can intervene.
As cooperatives, our (ad)vantage point is the principles we uphold, such as our commitment to democratic ownership and governance and our emphasis on social values. It also includes ecosystems and mutual relationships we have built, from the community level, to inter-cooperative networks, to partnerships between communities, cooperatives and public institutions.
In other words, what is realistically achievable right now for cooperatives is that we can tame AI, to some degree, aligning it with our cooperative identities as we find opportunities to intervene in the layers that make up this technology.
As for where and how to begin, rather than working in isolation, we must join forces with value-aligned mission-led organizations, and leverage an existing ecosystem of the social solidarity economy with shared interests and needs. Only through these partnerships can we hope to avoid the tragedy of fragmentation where some cooperatives become too isolated to be impactful and fall into a cycle of dependency, while others, having grown too successful, risk being eventually absorbed through corporate capture.
Allow me to link these lessons with a few observations that struck me when placing Tara’s and Kenzo’s works alongside the Bread Cooperative.
Though each project focuses on a different technology (AI and blockchain respectively), one common thread is their shared rejection of the binary assumption that any given technology must either advance the world or threaten it.
These projects hold a similar belief in the human capacity to collaborate and push back against narratives that tell us we have no alternative, therefore affirming the value of human agency over technological determinism and profit extraction.
Neither romanticizes the relationship between technology and humanity. Instead, they demonstrate that, in practice, technologies are not neutral, but neither are they fixed. Their impact and implications depend largely on our understanding of their underlying architecture and how we choose to engage with them.
If we treat technologies as something we simply adopt, we inherit their embedded assumptions. What these projects call attention to, by contrast, is the space in between, where human collaboration can shape the design, ownership, governance, appropriation, and cultivation of these technologies in ways that reflect and retain our social values.
For Bread Cooperative, it is through building blockchain-based applications that help collectives, cooperatives, and non-profits fund themselves on their own terms, and through experimenting with how grassroot financial mechanisms can be leveraged by blockchain technologies.
For Tara and Kenzo, as well as for other ICDE fellows whose works I am fortunate enough to come across, it is through empirically grounded research that strives to harness AI and its infrastructure to serve people and communities, and through their provocative questioning of how the entire AI stack could be structured in a radically different form. Solidarity AI is in the making.
Finally, I believe Bread provides one specific lesson to the question of sustaining non-extractive, cooperative-owned AI alternatives: solidarity technology needs solidarity funding.
Since its inception, Bread Cooperative has focused on addressing the funding problem that has long plagued the crypto industry. Typically reliant on venture capital and institutional investment, crypto projects have been pressured to launch products prematurely, prioritize maximum profits, and sacrifice community participants and users as mere “exit liquidity” for early investors.
To address this, Bread cooperative built tools like Solidarity Fund as decentralized financial mechanisms that enable non-profit and mission-oriented projects to remain self-sustaining and independent through community-led financial support.
Whether the Solidarity Fund truly lives up to its “solidaristic” name depends on how one defines the term. And it is important to clarify that I didn’t raise this point to suggest that cooperatives should blindly participate in Bread’s Solidarity Fund without a sufficient understanding of how blockchain and cryptocurrency actually work.
What is really at stake is this: what financial mechanisms can help mission-driven cooperatives that are willing to take part in our journey towards Solidarity AI to raise funds and sustain themselves without surrendering decision-making power and ownership, or being pressured into work that conflicts with their own values? What alternative funding schemes might allow cooperative alternatives to emerge in the context of Solidarity AI? And what kinds of financial infrastructure do they need to resist the pull of state and capital capture while remaining true to their cooperative principles?
Life under capitalism is full of contradictions. But resisting it is neither simple nor free from its own paradoxes.