The United Nations Office Geneva in the Palais des Nations. Shutterstock

The United Nations Office Geneva in the Palais des Nations. 

The World Doesn't Need Another UN AI Declaration. It Needs Architecture.

European officials release the “Tech Sovereignty Package” on June 3, 2026. Source

European officials release the “Tech Sovereignty Package” on June 3, 2026. 

Does Europe Really Have a Plan for Tech Sovereignty?

A neural network comes out of the top of an ivory tower, above a crowd of people's heads (shown in green to symbolise grass roots). Some of them are reaching up to try and take some control and pull the net down to them. Watercolour illustration.

How AI Keeps Europe Hooked on US cloud

People hold signs in opposition of xAI's use of gas turbines at their facility during a meeting of the Memphis and Shelby County Air Pollution Control Board, Monday, Dec. 15, 2025, in Memphis, Tenn. (AP Photo/George Walker IV)

Inside SELC's Clean Air Case Against xAI in Memphis

The undying sun hangs in the sky, as people gather around signal towers, working through their digital devices.

Digital Nomads Across Time by Yutong Liu & Digit. 

Imagining Broadband Policy of, by, and for the People

US Commerce Secretary Howard Lutnick listens to President Donald Trump’s speech during a news conference at the G7 summit, Wednesday, June 17, 2026, in Evian-les-Bains, France. (Jeanne Accorsini/SIPA via AP Images)

Commerce Eased Its Block on Anthropic's Mythos, But Major Questions Remain

(L-R) Congressional candidate Claire Valdez, Congressional candidate Brad Lander, Mayor Zohran Mamdani, and Congressional candidate Darializa Avila Chevalier raise their hands during the 'Our Team, Our Year' Get Out The Vote (GOTV) rally for local candidates ahead of this week's Primary Elections at the Kings Theatre in the Brooklyn borough of New York City, NY, June 18, 2026. (Photo by Anthony Behar/SipaUSA via AP Images)

Populist Candidates Need More Tech Policy—Reforming Tenant Screening Is A Good Start

The New UK Prime Minister Inherits a Social Media Ban. The Real Goal Is Safety by Design.

Tear gas surrounds federal law enforcement officers as they leave a scene after a shooting on Wednesday, Jan. 14, 2026, in Minneapolis. (AP Photo/John Locher, File)

Trump’s Crackdown on Dissent is Targeting Signal—and Zines

Split-screen image depicting contrast. On the left, a Rubik’s cube displays blue, red, and yellow sides, with a LinkedIn 'cr' AI-generation icon nearby. The background is divided between discarded electronic waste on a street and data centre cables, some connecting to the cube. On the right, the entire scene is repeated with inverted colors.

This is not a(i) game! by Laura Sofia Martinez Agudelo / 

When AI Agents Fail, People Ask the Wrong Question About Why

Harvard University graduates, top, celebrate during commencement exercises on the school's campus, Thursday, May 28, 2026, in Cambridge, Mass. (AP Photo/Steven Senne)

For Students, the Process of 'Becoming' is the Challenge No Chatbot Can Solve

Illustration of a person standing with their back facing the viewer, surrounded by large stylized eyes watching them from above. The person is in a spotlight formed by overlapping blue and gray geometric shapes, conveying a feeling of being observed or surveilled.

Who Really Controls Your Digital Likeness in the Age of AI Wearables? Not You.

Paper collage. A photograph of the Atomium in Brussels, cut into diagonal strips and reassembled over a grid of numbers.

Seeing More — Seeing Less by Anna Riepe & FARI / 

Google and Apple’s Anti-DMA Lobbying Strategy Goes All-in on Security and Privacy

A child holds up a smartphone, obscuring their face.

They Said I'd Feel Different About Free Speech as a Parent. They Were Wrong.

June 22, 2026—Andy Burnham with colleagues from the Parliamentary Labour Party in Westminster Hall at the Houses of Parliament in central London, as he returns to the House of Commons to take up his seat after winning the Makerfield by-election. (Yui Mok/Press Association via AP Images)

How Will Andy Burnham Handle Tech Policy as UK Prime Minister?

In this photo illustration, the Telegram logo is seen displayed on a smartphone screen with the Indian flag in the background. (Photo by Thomas Fuller / SOPA Images/Sipa USA)

India’s Telegram Ban Was Temporary. The Power Behind It Is Not.

The Anthropic logo is displayed on a smartphone screen placed on a reflective surface onto which a company branding illustration is projected on June 2, 2026. (Photo by Samuel Boivin/NurPhoto via AP)

Anthropic's Red Lines Are No Substitute for Public Law

BRUSSELS, BELGIUM - JUNE 9, 2023: The symbol of the European Parliament on the glass façade of the European Parliament. Close up

EU Lawmakers Press Commission on Child Safety as Debate on Age Limit Heats Up

Shutterstock Illustration depicting two megaphones with a breaking news text ribbon suggesting propaganda and fake news.

A Policy Playbook to Inoculate the Public Against AI Text Falsehoods

The image depicts a conceptual illustration of risk assessment. A large triangular warning sign with an exclamation mark is positioned prominently, symbolizing caution or danger. A business professional in a suit, holding a magnifying glass, is inspecting the sign closely. The scene suggests careful examination or investigation of potential risks, with a neutral background featuring minimalistic clouds to keep the focus on the subject.

The EU's AI Transparency Code of Practice, Explained

is published in partnership with the AI Now Institute.

In our previous iteration of this series, we talked about European AI start-ups that operate at the application layer — and why betting on certain versions of applied AI may further entrench Europe’s dependency on a handful of United States firms.

The obvious counterexample to start-ups like the Swedish vibe-coding company Lovable is Germany’s DeepL. Launched in 2017, the company develops its own, proprietary models at a 

 rather than matching the training-data volume Google built up over years of web crawling. DeepL is more than a European AI success story — it 

has outperformed Google Translate in accuracy assessments

 — and it made GDPR compliance, along with processing customer data on its own servers in Germany and Iceland, a core part of its public narrative, and a main selling point for privacy and sovereignty-conscious users.

That was true until April 2026, when the Cologne-based company informed its paying customers that it was entering a partnership with Amazon Web Services (AWS) and would no longer process data exclusively on its own servers. This announcement was publicly 

 as a deliberate decision and, at least implicitly, as a betrayal of the company’s commitment to Europe.

From a strategic perspective, it might seem puzzling: Why would DeepL, a company that has made sovereignty its key selling point, end up partnering with AWS, one of the (perceived) three horsemen of Big Tech?

In this piece, we argue why European AI Sovereignty is intricately connected to its existing dependence on three US cloud providers: Amazon, Microsoft, and Google.

As a result, each of Europe's attempts to build its way out risks deepening the very dependence it aims to cure. By boosting demand for AI, especially at the frontier, it also boosts demand for proprietary models bundled with the dominant cloud providers. By building gigafactories, it trains models that then run on hyperscaler infrastructure. And by building its own compute for inference, it risks subsidizing capacity for the leading US AI firms, the only customers large enough to fill it.

Hyperscalers own the distribution channels

Contrary to public opinion, DeepL’s decision can be understood through the brutal, structural reality of a highly concentrated global AI stack: DeepL’s enterprise customers (think multinationals) want the same service everywhere. AWS operates globally and can provide low-latency, real-time performance that no European cloud providers can match. Especially as the company is 

 to real-time voice translation, the feasibility of serving AI models globally from a data center in Southern 

 has become unfeasible. For an ambitious company with an IPO lined up for this year, loading the balance sheet with debt to facilitate its own data center buildout in a historically tight market would likely have faced resistance. Lastly, given the business need to compete with Google Translate's ubiquitous presence, access to AWS's end user market and distribution channels likely drove the decision: Once DeepL is listed on AWS Marketplace, enterprises can acquire, deploy and manage DeepL directly within their existing AWS infrastructure through a trusted procurement platform.

All of these factors point to the pervasive monopolies of the global computational infrastructure that pull European companies into their orbits, rather than the companies’ commitment to Europe.

Europe's AI industrial strategy since 2025 has been built around expanding the continent’s ability to train frontier models: centralized, frontier-scale training capacity designed to train, post-train, and evaluate the most complex models. This requires highly specialized compute infrastructure, equipped with (often Nvidia’s) latest AI chips. This framing is politically convenient, as it enables the leveraging and pivoting of existing public high-performance computing clusters (EuroHPC) towards the AI race. Recent AI Factories and Gigafactories are descendants of these initiatives, now facing 

 due to emerging funding gaps, elusive demand and difficulties in governance.

But even if these questions were solved, a more serious problem looms – widespread AI adoption doesn't automatically generate demand for one-off gigantic training. It generates demand for serving inference: responding to individual user requests one at a time, as they arrive, with low latency and high availability, like a user typing into a chatbot and expecting an immediate answer. Inference also requires high-capacity (though not the latest) GPUs, but is a fundamentally different kind of infrastructure than training. It is distributed, elastic, close to users, rather than centralized and connected. This leads to what we have 

— the tendency to fixate on training runs that are conducted on public computational clusters while ignoring the private compute ecosystem in which those models will eventually live and produce value.

Why does mainstream demand for AI products benefit hyperscalers

Now, the obvious question for European policymakers is how to ensure that models run on European cloud infrastructure. There are three reasons why even a huge demand for AI inference doesn’t mean that this demand can (currently) benefit European providers.

Proprietary models on hyperscaler infrastructure

What is worth reiterating for a policy audience is that AI models come in two types, and the difference decides who is allowed to run them. With an open-weight model, the developer publishes the model itself for download, so anyone (i.e., a startup, a university, a European cloud provider) can run a copy on their own infrastructure. Alibaba’s Qwen, Meta's Llama and France's Mistral are open-weight. The most capable models remain proprietary: Anthropic's Claude, OpenAI's GPT, Google's Gemini, even as open-weight models like DeepSeek and Qwen have caught up with the frontier tier.

Proprietary model makers distribute their models on the cloud providers they depend on for capital and compute – Amazon AWS, Microsoft Azure, and Google Cloud – largely bypassing European providers. 

SAP’s “sovereign” OpenAI for Germany

, for instance, also runs on Microsoft’s Azure cloud. This is also due to the fact that AWS, Microsoft, and Google Cloud have struck circular spending partnerships with the largest AI labs, to give them discounted computing costs (in exchange for either guaranteed generous or ownership stakes). In a highly concentrated AI market, where most demand is for proprietary AI models, more AI adoption means more demand for models run on hyperscale infrastructure. This creates a feedback loop: the money and data received from run models can be used to train the next generation of models.

All this puts European cloud providers in a bind. As they are not able to offer proprietary frontier models, they're left serving open-weight models. However, the largest open-weight models are very expensive to run, as they require providers to reserve specialized GPUs for that workload alone. Europe's main cloud providers have started offering large open-weight AI models, but mostly older or compressed versions, and usually not the latest. The clearest case is IONOS, whose largest model, Llama 3.1 405B, last led the field in mid-2024. Scaleway is furthest ahead, offering a model that was near the frontier in late 2025. Across all of them, the genuine cutting-edge open models of 2026 are still largely unavailable. (Accurate at the time of publishing; these catalogs change quickly.)

The hyperscalers AWS, Google and Microsoft already own their GPU fleets, which they bought for training and running the proprietary models they serve to their enormous, existing and captured customer base. As a result, their risk of building inference capacity is low. In the worst case, the company can spend the compute on serving their internal needs. A European competitor has to build the fleet from scratch and carry the risk of it sitting idle until demand arrives.

That's the chicken-and-egg problem, and the reason every European cloud provider that offers AI as a service focuses on smaller and medium-sized open-weight models. The choice is constrained on two ends: open-weight because the frontier proprietary models simply aren't available to host on someone else's infrastructure, and smaller because such models have a far lighter memory footprint and can be served on fewer GPUs and other specialized chips needed for inference. All of this is shrinking the fleet you have to bet on before demand materializes. But at the same time, the lack of competitive models is seen as a problem.

No open weight demand, no frontier serving in Europe

There is simply no demand (yet) that justifies the investment required to serve frontier models at scale. This also explains why European initiatives that are designed to expand (public) compute for inference end up partnering with large AI companies as anchor clients, as Leevi and Boxi have written in the case of Italy’s Apto data center - the only data center investment currently in the infrastructure portfolio of the European Investment Fund (EIF)

The only private actor that currently has a strategic incentive to diversify inference compute in Europe (and globally) is Nvidia. Its problem is concentration: a handful of US firms account for the bulk of its data-center demand, and those same firms — Microsoft, Amazon, Google, Meta — have each developed their own chips, predominantly for inference serving, but increasingly also for frontier training. That is precisely why Nvidia has had a 

 a class of so-called neoclouds, AI-specialized compute providers, in Europe. Dutch 

 are notable examples, with Mistral's expansion into computing infrastructure as an emerging competitor in this space. Through investing in companies, providing priority access to the latest chips, and guaranteeing demand through binding purchase commitments known as offtake 

, Nvidia entrenches itself as the computational infrastructure and diversifies its buyer base. However, even many of these Nvidia-backstopped computing providers 

 suppliers of compute for the American tech in search of a sufficiently high demand for their expensive hardware.

What Europe gets wrong about AI and the cloud

The reality of the market clashes with three key assumptions behind Europe’s AI industrial policy strategy. From the 

, the focus has primarily been on (moderately funded) attempts at building compute for frontier model training and boosting AI adoption. This focus has been based on the following assumptions:

That investing in compute for frontier-model training and European model makers will automatically lead to less dependency on dominant firms;

That European AI companies can make a conscious choice to avoid US hyperscalers;

Perhaps, most importantly, that cloud is a generic capacity for which there is a generic demand and supply.

The last assumption in particular is a gaping hole in the proposed Cloud and AI Development Act (CADA), Europe’s latest legislative proposal aimed at tripling the continent’s cloud capacity, while ensuring a small fraction of public sector demand goes to European providers. But that ignores that there are different supply and demand logics behind different cloud services. European cloud providers vary widely: Germany’s Hetzner, for instance, is very competitive on price, with a strong developer following and dedicated GPU servers, but doesn’t offer an "AI-as-a-Service" (AIaaS) platform with pre-built APIs for models. France's Scaleway, for instance, runs a genuine managed-inference / model-as-a-service layer, not just GPU rental, yet still lacks AWS’s global scale and distribution channels.

Even its flagship customer, Mistral, is now building its own infrastructure with Nvidia support instead of scaling on Scaleway. Schwarz Digits, meanwhile, the digital arm of Schwarz Group, the retail conglomerate behind Lidl, wants to become a European hyperscaler.

Demand, too, is not only for raw compute and storage, but for the higher-level services bundled with them: the managed services, as well as the surrounding ecosystem and tooling. Increasingly, demand for cloud services includes demand for proprietary AI models. And much of this “demand” is much less of a fungible consumer choice than the result of contractual lock-ins and other practices currently under 

. As a result, even under the perfect interoperability conditions, switching may not always be feasible.

As we have seen in the response to DeepL’s choice to partner with AWS, there’s a tendency to misread a structural pull as a moral choice. No European provider could have offered what DeepL needed. But the conclusion to draw from this is not that AWS simply offers a better product. By controlling the distribution and the demand (as well as locking in existing consumers), the hyperscalers can shape the strategies of actors downstream.

Europe’s dependence on three US cloud providers is now widely understood in EU policy debates, and so is Europe’s dependence on a small number of US companies for access to large-scale AI models.

What remains much less understood is how cloud concentration and AI dependence relate to each other: By boosting generic demand for AI (especially at the frontier), Europe risks boosting demand for proprietary models that come bundled with dominant cloud providers. By building gigafactories, Europe risks training models that then run on hyperscaler infrastructure. And even by building European compute for inference, Europe risks subsidizing infrastructure for the leading US labs, the only customers with enough demand to fill it.

AI sovereignty is not separate from the existing cloud market dynamics Europe has failed to grapple with. It is intricately connected in ways that simply boosting supply and demand separately can’t solve.

It’s not all doom and gloom, but the reality of the AI market leaves policymakers with some uncomfortable choices. We will discuss these in the next piece of this series.

Somewhere in Geneva next month, a diplomat will declare that artificial intelligence must benefit all of humanity. That statement is easy to make — and nearly impossible to disagree with. The harder question, and the one that will define the United Nations’ first 

, is whether fine words this time produce the architecture to back them up — architecture built not only by the powerful, but also on the knowledge, participation and agency of the global majority.

Governments, civil society, researchers, private enterprises and technical experts will gather to discuss some of the most critical issues of our time. The timing is critical because frontier technology, including AI, is no longer an abstract technology policy issue. It is reshaping labor markets, public services, education, information ecosystems and the agency and cognitive abilities of people across diverse communities here in Asia.

The institutions currently shaping AI governance remain concentrated in a small number of wealthy countries, powerful companies and well-resourced research environments. For many countries in Asia and the Pacific, AI systems are arriving faster than the regulatory, technical and civic infrastructure needed to understand, evaluate or challenge them — and well before communities have built the resilience to absorb harms that are already forming. As our region’s policymakers simply try to keep pace with foreign AI technology, we’ve barely moved the needle on discussions around catastrophic and existential risk.

That is why the UN Dialogue remains critical for humanity. The UN is the only forum where every member country has an equal seat, and where civil society, affected communities and technical experts can press for governance that reflects more than the priorities of dominant markets, powerful governments and wealthy private interests.

But the Dialogue will only succeed if it becomes more than an annual convening bringing together the familiar, privileged voices. It must center the world’s majority and produce the architecture needed to turn shared principles into practical governance.

First, the Dialogue must be judged by follow-through, not participation alone. A successful first session should close with a limited and prioritized set of workstreams, named and regionally balanced stewards, a clear reporting cycle, and a formal route for stakeholder evidence to enter deliberations. In this vein, the 

Without these basic structures, the Dialogue risks becoming a symbolic event: widely attended, diplomatically discussed and generally ignored in substance. We don’t need another forum that produces de minimis consensus language while leaving each country to interpret it alone. It needs a recurring mechanism that helps stakeholders alongside governments identify what is working, where implementation is failing and what support is needed to close the gap.

Second, capacity-building must be treated as the foundation of AI governance, not as a secondary developmental concern. Safe, rights-respecting and accountable AI governance cannot be implemented without trained regulators, judges, parliamentarians, auditors, procurement officials, incident responders, researchers and civil society groups. Nor can it succeed without practical infrastructure: policy observatories, testing tools, audit pipelines, harm-mapping systems, language-specific evaluations and institutionalized routines.

This is especially urgent in Southeast Asia and across the Global Majority. AI systems are often designed in a few Northern jurisdictions (often with weak or absent regulatory safeguards) but deployed globally. When a regulator in our countries cannot evaluate a frontier model, a procurement office cannot assess algorithmic bias or a judge cannot adjudicate an automated decision, the result is not merely a local capacity gap with localized implications (that can be ignored by the privileged in the North). Rather, substantively it becomes a detection, response and remedy problem for everyone.

Capacity is also a matter of sovereignty. Governments that lack domestic expertise cannot negotiate confidently with frontier-model providers from an informed baseline. They may find lawmaking effectively outsourced to foreign consultants or corporate actors.

Third, the Dialogue must recognize Global Majority expertise and lived experience as governance knowledge. Too often, global AI debates are framed by commercial incentives and national-security concerns generated in a narrow set of institutions and geographies. Yes, these perspectives matter, but they are also incomplete. They can overlook harms that are harder to quantify: institutional degradation, weakening of community bonds, cultural and linguistic exclusion, exploitative labor practices, gendered abuse and erosion of public trust and information integrity.

People experiencing AI harms are not merely “stakeholders” to be consulted after policy options have been drafted. Workers whose tasks are being redesigned, women and girls targeted by AI-enabled abuse, children and other vulnerable people exposed to unsafe AI companions, migrants and gig workers, people with disabilities, gender and sexual minorities, low-income users and local-language researchers all hold knowledge that governance systems need in order to work.

Lastly, the Dialogue should create legitimate space to discuss catastrophic and existential risk. These risks should not be treated as speculative concerns reserved for a handful of wealthy governments, frontier labs or Northern research institutions. Compute is scaling rapidly, and evaluations already show AI systems completing increasingly long and complex tasks once requiring focused human expertise. The Dialogue can help ensure these questions are addressed transparently and inclusively so that we can define which capabilities require collective oversight, restraint and coordination.

The Dialogue should therefore change how AI governance knowledge is produced. Risk taxonomies should incorporate harms identified by affected communities and Global Majority researchers. Regional research and language-specific evaluation work should be commissioned and cited alongside work produced by frontier AI developers. Dialogue outputs should be reviewed not only for whether they consulted diverse groups, but for whether those groups’ knowledge actually shaped the conclusions – and improvement recommendations. And the Dialogue must create the space needed to have very serious conversations about catastrophic and existential risks in a way that private interests cannot dominate the narrative. Ultimately, the evolution of AI and frontier technology is a public interest issue and the tools that result should be treated as public goods.

The Dialogue should therefore establish a standing, regular and inclusive mechanism to surface and deliberate on catastrophic and frontier risks. This means a dedicated workstream with regionally balanced membership, a mandate to commission independent technical assessments and a formal process for Global Majority governments and researchers to shape the risk frameworks themselves, rather than inherit them from frontier labs or Northern institutions.

The UN Global Dialogue has a rare opportunity. It can help align AI governance around shared values: human agency, dignity, rights, accountability, environmental sustainability and equitable prosperity. It can advance pro-social AI. But these values will not implement themselves. They require institutions, evidence, resources and mechanisms that allow countries and communities to act.

The test for Geneva is not whether participants can agree that AI should benefit humanity — that bar is too low. It is whether the Dialogue produces the architecture to make that promise credible, and whether the communities with the most at stake helped design it.

The Dialogue's value will not be measured in declarations. It will be measured in what gets built afterward, by whom, and for whom. That is the standard the Global Majority should hold it to.

Audio of this conversation is available via your favorite podcast service.

In June 2024, the Greater Memphis, Tennessee Chamber of Commerce 

 a deal to host a substantial new computing facility. xAI built its "Colossus" data center in an old Electrolux factory. The facility was brought online at record speed, and almost immediately drew questions about transparency, air quality, water use and how a project of this scale had moved through local government so quickly.

Two years on, the story continues to expand alongside the company’s growing footprint, with a second campus, Colossus II, across the state line in Southaven, Mississippi; a contested gray water recycling plant; an ever-rising count of gas turbines; multiple lawsuits; and communities in South Memphis and beyond still pressing for straight answers.

In multiple ways, the story is a harbinger, a case study for communities across the country now reckoning with the arrival of hyperscale data centers.

One of the entities leading the way in Memphis is the Southern Environmental Law Center (SELC). The SELC has been a partner 

, gathering evidence on the company's unpermitted gas turbines and pushing state regulators to hold the company accountable under the Clean Air Act. To learn more about SELC’s case and what it means for the broader fight for transparency and accountability amid the AI infrastructure boom, I spoke to 

, senior attorney and data center project leader in the Tennessee office of the Southern Environmental Law Center.

A note since we recorded: On June 15, the Department of Justice 

 that shutting down the turbines powering Colossus 2 would threaten national, economic and energy security. The department claimed Grok is one of only four AI models supporting mission-critical operations on classified networks, and a Defense Department declaration tied the model to military operations, including recent strikes against Iran. The DOJ, xAI and the state of Mississippi all asked the court to dismiss the suit. SELC 

 the government’s argument “an unprecedented attack on the public’s ability to defend themselves from illegal pollution.”

This is the second of three episodes on the situation in Memphis and Southhaven, and what it portends about the future of the AI infrastructure boom in the United States and the political debates and power struggles surrounding it. Listen to the first—an interview with Memphis Commercial Appeal reporter Neil Strebig—

What follows is a lightly edited transcript of the discussion.

Amanda, I think I have to ask you to do two things for our listeners. One, give them the quick one-two about what SELC is and what you do looking after data centers there.

The Southern Environmental Law Center is a law and policy organization that is rooted in the South and we work across a full range of environmental issues. Our theory of change is that place matters, that people care about the places that they live. And so we have attorney boots on the ground in all six states where we work. We work very closely with communities in Virginia, North and South Carolina, Georgia, Alabama, and Tennessee. And our advocacy, even though it is rooted in the South, really pushes issues, elevates issues at the national level as well because we are able to tell the story of how policies affect people on the ground.

In our data center work, I think is a perfect example of that. We have been working alongside communities on local land use issues for four decades. Also, we have been in public utilities commissions at the state level for four decades. So we really have those relationships that allow us to be effective on the ground, but also to drive policy at both the state and the federal level. And in my role as the data center project leader, I help to develop strategies and guide our attorneys on data center matters across the region.

So it's been two years since xAI announced that it would build the world's largest supercomputer in Memphis after closed-door negotiations with the Economic Development Authority there. It is very shortly after that community members spotted gas turbines appearing and raised concerns about a lack of permits. Walk us through the Memphis story. When did SELC get involved? When did this come to your attention?

Justin, I've been involved since day one of the xAI issue in Memphis and really it again highlights the importance of working with communities over the long term because we had been working in Memphis for many years and we have a coalition of local environmental justice and environmental organization that meets weekly about issues that happen or that come up in the community.

And so we had gotten wind of the fact that there was a big announcement coming. And when it happened, our local community partners were ready to go. They issued a statement on day one, the day of the announcement, raising all the important questions that weren't being addressed by the Memphis Chamber and the mayor of Memphis.

Our local partners immediately issued a statement asking, "Where is the water going to come from? Where is the energy coming from? How much water and energy is xAI going to use?" And that initial statement I think really immediately changed the conversation locally from a victory lap for the economic development folks to a conversation about how is this going to affect the communities on the ground in Memphis?

And take us through the claims in your lawsuit. Where is the lawsuit at the moment? What are the kind of key claims that you're pursuing?

I think it's important to take a step back before we get into the current lawsuit because, as you pointed out, this is a situation that's been unfolding over the past two years and there have been multiple phases, and I think it really highlights the lack of transparency around data center development and the way these projects can kind of metastize and grow over time.

So when we initially heard about xAI, it was the Colossus 1 data center and the idea was that it was going to ultimately be a data center of 150 megawatts and that it would be served by the grid. But as you pointed out earlier, that scenario quickly changed and we learned that the grid connection was not going to happen fast enough for xAI's timetable. And so they set up unpermitted, uncontrolled combustion turbines. Initially it was 18 combustion turbines. At that point we raised concerns actually with the local air permitting authority, the Shelby County Health Department, and basically sent a letter and said, "Hey, these guys are running what is basically a power plant. They don't have a permit. You need to ask them to get a permit or require them to get a permit."

And the health department at that point said, "We don't think this is our responsibility. We think it's EPA's responsibility." And they basically punted to EPA. That was in August of 2024, so just before the presidential election. And initially it appeared that EPA was going to weigh in and they said they were going to issue a regulatory interpretation. And after the election happened, EPA went silent. At that point, we were not getting information from xAI, we were not getting information from the local air regulator, and meanwhile, these combustion turbines were pumping pollutants into the air 24/7. As far as we knew, they were operating 24 hours a day, seven days a week.

And that's particularly concerning because we're talking about air pollution in Southwest Memphis. It's a known air toxics hotspot and the whole Memphis area is not meeting minimum federal standards for ozone pollution, which is also known as smog. So we were really concerned about the public health implications of these turbines. So we took matters into our own hands. We've actually worked with an organization called SouthWings that provides air flights for community groups that are investigating environmental damage across the South. And through that flight, we discovered that there were not just 18 combustion turbines anymore. There were 35 and we put that information out publicly. The mayor initially said that xAI told him, even though there were 35 on site, there were actually only 15 operating. We did another flight with a thermal camera and showed that 33 of the 35 were operating.

And at that point we decided that we couldn't wait for regulators to act any longer. And on behalf of the NAACP, we sent a 60-day notice to xAI. That notice basically said, "You have constructed and are operating a major source of air pollution without getting a permit," and asked them to cease operating until they got a permit. At the same time, they had applied for a permit to operate some different turbines on a permanent basis. They claimed that the ones that they were operating without a permit were temporary. And so we were also participating in that air permitting process.

This is all for Colossus 1. After we sent our 60-day notice, xAI did end up removing all of the temporary turbines and got their permit. So in a sense, on some level we won, right? They got rid of the temporary turbines, the ones that didn't have any pollution controls. They replaced them with permanent turbines that did have the pollution controls that have to do reporting and all the things that you would expect from a power plant.

But in the meantime, xAI had also grown. Somewhere along there, and I can't remember exactly when, they announced that they were actually going to grow Colossus 1 to 300 megawatts. Eventually all of that power, all the 300 megawatts would come from the grid and they would use these combustion turbines as backup power. But in the meantime, until they could get fully grid connected, they were going to keep running first the unpermitted turbines and then later when they got a permit, the permitted turbines.

So after xAI's permit was issued and they had removed all the turbines, we actually filed a permit appeal for Colossus 1. We challenged not just the permit, but also the health department's decision to adopt a policy of not requiring permits for what they called temporary turbines. Unfortunately, we filed that permit appeal on behalf of NAACP and Young, Gifted & Green. We had a hearing. There were hundreds of people there as there had been at the original permit hearing, and ultimately the air board determined that our case was moot. And so that question of whether they could not permit temporary turbines was not answered, but shortly, actually less than a month after we had that hearing, EPA actually issued a new rule about these turbines and expressly said, "We have always treated temporary turbines as stationary sources that require permits."

So we thought, okay, now we've got a clear statement from EPA that these things need permits, but in the meantime, xAI had expanded again. They had acquired land also in Memphis at Colossus 2 and they had also acquired a site in South Haven, Mississippi, just across the border from Colossus 2 that is the site of a former power plant. And it turned out that when they removed all of those turbines from Colossus 1, they just wheeled them across the border and set them up at this former gas plant site in South Haven, Mississippi. And they told a local gathering of the business community that they were planning to copy and paste what they had done at Colossus 1, at Colossus 2 at Southaven.

Well, I guess I'll kind of jump in there. So at this stage we've got now multiple facilities. On some level, it sounds like we know xAI was keen to get up and running as fast as possible. Effectively, they moved very, very quickly. The authorities effectively weren't able to keep up on some level. All these processes that you're referring to happened after the fact. They get Colossus 1 up and running and powered and then they recognize that they want to build more capacity so they go to South Haven, go to Colossus 2, et cetera. There's maybe a little bit of almost like a kind of arbitrage of what's going on with these gas turbines, but where are we at right now? I mean, are they still powering South Haven with turbines or what's happening there? Have they succeeded in developing more of a power source? I know that was part of the proposition at one point, that they were going to build a more long-term source of energy.

Right. Great question, Justin. They are still operating unpermitted turbines at the South Haven gas plant, and it's worth stating that, okay, so Colossus 1 is 300 megawatts. Colossus 2 is 1,200 megawatts, so a much larger data center. And that data center is currently being powered by permitted turbines that are, some of them have some controls on them, but overall they are similarly not, they don't have the types of controls that they would need to have if they were permitted.

In terms of where things stand right now, we have filed a... So we once again sent a 60-day notice on behalf of the NAACP, same kinds of claims about constructing a major source without getting a permit. And we have filed litigation in the Northern District of Mississippi. We have also filed a preliminary injunction motion in that case and that is set for hearing in August. So basically two years after the initial announcement of xAI coming to Memphis, we will be in court litigating over these unpermitted turbines in August.

I understand with that preliminary injunction request pending, you also now expect the Justice Department to get involved. Are we expecting the federal government to effectively attempt to intervene here?

They have indicated that they are considering intervening and we should learn more about that in the month ahead.

I really can't talk about that aspect of the litigations because it's pending, but I will say in addition to... We talked about xAI copying and pasting from Colossus 1. There is a difference here, which is that xAI is not intending to ultimately connect to the grid for Colossus 2. So instead, they proposed their own permanent power plant of sorts, but it's not a combined cycle gas plant, which is what you would expect for something of this scale like 1.2 gigawatts is a lot of power. It's enough to power a mid-sized city. Instead, it's a collection of multiple types of combustion turbines, including some that are just refurbished jet engines. And they obtained a permit from MDEQ, the Mississippi Department of Environmental Quality.

And so we have also filed a permit appeal for that installation of turbines, which again is separate from the unpermitted ones that are operating currently. They're on the same site, but they're two different sets of turbines. And that permit appeal is, we are in the process of... We're first of all, I guess, representing NAACP, Young, Gifted & Green and Safe and Sound Coalition, which is an organization of local neighbors that live near the plant. That is set for hearing in November of this year. So we have challenges to the temporary installation and the permanent installation of combustion turbines at this South Haven site.

I want to ask you a question about this idea of speculative demand. I know you commissioned a report on the idea that projected data center energy demand, I suppose data center demand generally may be wildly overstated. Can you talk a little bit about your findings there, how it matters for the cases that you've got in front of you right now?

Sure. So I think we've actually commissioned two different reports, and 

 ,and that one really looked at if you add up all of the demand projections for data center development, it would outpace the ability of chip manufacturers to provide chips to run those data centers. So it really calls into question whether what is being proposed by the data center developers is really feasible even from their own technological standpoint. And you're starting to hear tech companies talk about that, like the pressure on the chip market.

And once we had that report, we decided to look more specifically at our own utilities in our region and the demand forecasts that they were using in their energy planning. And when we took a look at that, once again, when you compared what our utilities were projecting with third party estimates of potential demand over the next 20 years or so, they didn't match up. And there's a certain amount of course, data center developers are shopping around, they're getting in the queue in multiple jurisdictions, but they're ultimately only going to develop a data center in one of those jurisdictions. So that's part of it.

The other part is that our utilities are regulated, mostly vertically integrated monopolies. They make their money by building new shiny things. And so they have historically had an incentive to overbuild new capacity and this has just kind of put that on hyper-speed in terms of the amount of investment.

The problem, of course, is that if all of that demand doesn't materialize, who is going to be left paying for these stranded assets? And right now, the answer, in our region at least, is largely everyday rate payers are the ones that are going to be left holding the bag if we overbuild. And we are almost certainly overbuilding because ultimately there's going to be... All this data center development is not going to come to fruition, and some of it that does come to fruition may fall by the wayside over time.

So you mentioned that SELC has been working on these types of issues for decades. If you look back at the history of industrial development in the South, maybe more generally, is this a kind of precedent-breaking industry or does this look like other things that your organization has seen in the past? I mean, clearly the scale is historic. I mean, the amount of many, many, many hundreds of billions, trillions that are being invested in data centers across the country and the world is one of the largest infrastructure projects, I assume, that we can point to as a singular effort of the species on the planet over some period of time. But I don't know, are there patterns you see or ways that this breaks the pattern?

I think this is breaking the pattern. As you point out, the scale and also the pace of development. I mean, and xAI is a perfect example of this, right? They came in and built a data center in 122 days and they have expanded that massively over the past two years. And we're talking again about the type of infrastructure that normally would take years to develop and a power plant that provides power for... that would otherwise be providing power for a mid-sized city and is just being operated to serve this one company. The amount of water that the xAI data center requires is also hugely significant in terms of relying on a community's fresh drinking water source.

So I think we're seeing in particular the systems that we use to process development applications and go through permitting and all of that are not set up for something of this scale. These are something that economic development... There's been a use of NDAs, non-disclosure agreements in economic development for a long time, and they've always been problematic, but they are even more problematic when you are talking about impacts of the scale that we are talking about.

And what we hear from communities over and over again is we need to have a seat at the table when we're talking about impacts that could have these profound effects on our community, on our shared resources like drinking water, like air quality, like energy usage, all of these systems that we all rely on together, and our decision-makers are making these decisions kind of behind closed doors without community input. And so I think I really think about the data center rush as pushing these systems to the breaking point and it's the reason why we need better policies.

And that's what you're starting to see across our region and across the country. It's starting at the local level because that is where people are feeling the impacts most directly. And I literally this morning got a picture from my colleague in Birmingham of 200 people waiting to get into the Birmingham City Council meeting where they are deliberating over common-sense protections for what are the terms on which we are willing to allow data centers into our community. You see communities across the Southeast adopting moratoria. They're pressing pause. They're saying, "We know our systems are not set up for this and we really need to have a conversation as a community about whether and on what terms we're going to allow data centers to be part of our landscape, to be part of our water system, to be part of our electricity system."

So I do think that is happening. We're also starting to see conversations happening at the state level. There was a lot of data center legislation proposed. I know Tech Policy Press just published a great article that kind of summarized the state of play at the state level. And it's early days in terms of what the state level protections are going to look like, but it is what we're seeing on the ground that is driving those conversations. And same thing at the federal level. We're starting to see some legislators proposing legislation to address the impacts of data centers, but it is early days, but it is very much being informed by the experience of people on the ground and the concerns that are being raised by local communities.

Does seem to me like Memphis also kicked off in many ways the national news cycle on this, that it was one of the most prominent examples, most severe examples I think in many ways the kind of mixture of variables in that particular situation lent themselves to lots of coverage and the personalities involved, the companies involved, the nature of the harm to the local community there was so apparent in many ways. Also, a lot of really amazing activist voices, folks from the civil rights community and others who I think did an enormously good job of getting out the word beyond Tennessee certainly about what was going on there. Are there other places in the South now, you point to Birmingham, are there other places that you're working in that you think may be other points of national prominence over the next months or years as this conversation carries on?

I think we are really seeing the data center issue across the spectrum of rural, urban, low wealth, wealthy, conservative, more liberal communities in our region. Memphis is obviously an urban predominantly lower wealth, Black community, but we have seen and are continuing to see a lot of data center development in more rural areas in our region as well. I don't have a crystal ball about which one is going to become the most highly visible, but I will say that I think that communities across the southeast are really looking to the model of community activism that Memphis provided.

And you mentioned the Civil Rights Movement, which of course Memphis is a center of civil rights history, and the same sorts of community organizing principles that have informed civil rights advocacy as well as the environmental justice movement are being used by communities of every type across the country right now to push back against data centers.

So our clients at the NAACP actually convened a meeting in August of, I think it was in August of 2025 where they developed a frontline framework, basically it was informed by the 17 principles of environmental justice and the Jemez Principles of Democratic Organizing, but it applies those specifically to the data center context. And I would encourage anybody to go look at that frontline framework in terms of how the community organizing is happening.

I will point to another way in which xAI was unfortunately influential and then I'll tell you about a situation that's happening right now that illustrates this point, which is basically this idea of standing up their own power source onsite or nearby rather than waiting to get connected to the grid. President Trump has endorsed this idea of developing fossil-fuel-powered onsite or dedicated generation. I call them popup power plants and we are unfortunately starting to see more of those in our region. And so then you're talking about not just the impacts of the data center itself on site, but also all of the attendant air pollution and the same concerns that we have had at xAI.

And a recent example of this is happening, is unfolding right now in Spartanburg, South Carolina, a very rural, very conservative community where a data center initially set up with 50 megawatts of onsite power and recently announced that it is going to increase that to 450 megawatts of onsite power and the community is really getting organized and engaging, is very opposed to this idea that they're going to be hosting this massive data center and a power plant onsite to serve it. So I do think the popup power plants and more of the rural data center development are kind of a big theme of what we're seeing right now.

I think the primary thing that I would want folks to take away from xAI is that it is a cautionary tale for communities in terms of what happens when communities don't have a seat at the table, right? You end up with a project that is not in line with the community's vision or values and the importance for local leaders and policymakers of making sure that communities are involved. And so we have a guide at SELC called Getting It Right: Local Approaches to Data Center Development that talks about making sure that your local zoning ordinances are allowing for that opportunity for public engagement and input early in the process before the data center locates and gets up and running.

The other thing that I would just want to highlight for folks is that although data centers are posing a huge threat to Southern communities, they are also providing a huge opportunity that we see people taking advantage of. And that is the opportunity to get involved in the democratic process, to show up to public hearings, to show up to your city council meeting, to advocate for what you want. And we're really seeing that engagement again across the political spectrum, people demanding more and better from their local leaders to do the work that government is supposed to do in balancing the needs and interests of industry with the needs and interests of the community.

So I really take a lot of hope for democracy from the data center work because I am seeing such profound levels of engagement from people all across the South on this issue. As you pointed out, it started in Memphis. It started with a community that is deeply rooted in community organizing and advocacy and civil rights ensuring that people get a seat at the table, and that's the model that we're seeing play out across the country and it's pretty inspiring.

Access to affordable, reliable high-speed internet is a prerequisite for nearly every part of modern life, from finding work and finishing schoolwork to seeing a doctor or staying in touch with family. Yet millions of American households remain stranded on the wrong side of the digital divide.

The Blueprint for Equitable Digital Participation

," a report released in May from Public Knowledge, UnidosUS, and the National Digital Inclusion Alliance. Rather than beginning in Washington policy circles, this report centers the lived experiences of low- and moderate-income households to find out what's actually standing in their way and what should be done about it. Justin Hendrix had the chance to dig into the findings with the report’s authors: 

, broadband policy director at Public Knowledge, and 

, senior civil rights policy advisor at UnidosUS.

My name is Alisa Valentin, and I am the Broadband Policy Director at Public Knowledge.

And my name is Claudia Ruiz, and I'm the Senior Civil Rights Policy Advisor at UnidosUS, the nation's largest Latino civil rights and advocacy organization.

Pleased to talk to the two of you today and see you again. I just saw the both of you in Washington DC recently and we're going to talk a little bit about this report you've produced, “

” that has just came out, and it is a work product of Public Knowledge and UnidosUS and also NDIA, the National Digital Inclusion Alliance. How long did this tome take to put together? This is 70-odd pages of material, including all the citations, as that's quite a lot more.

Yeah, it took a couple of years. We started this project in 2024, so this is kind of a part of Public Knowledge's ongoing effort to ensure that we're centering equity in all of our work through our partnerships on the national level and on the local level in the way that we advocate and what we advocate for as well. So in 2024, probably in the spring, I approached Claudia, whose organization has a ton of partners on the ground, as well as the folks at National Digital Inclusion Alliance, and I was essentially like, "Let's get out of DC. Let's talk to people in communities. Let's elevate their experiences." And I was super happy that Claudia immediately said yes and NDIA immediately said yes, and so it took about two years. Now, I don't know Claudia if you have anything you want to add.

No, and I think the way that Alisa narrates it is exactly how it came together, where I think the one thing that our three organizations, despite the fact that we all have different missions, different areas of expertise, and even different sort of network or play in different coalition spaces, the one thing that unites the three of us is really our focus and emphasis on trying to channel authentic community voices all the way from on the ground up into Washington DC. And making sure that it is those authentic voices, it is the very constituents and voters themselves, that are actually communicating the kind of needs and solutions directly to law and policymakers. And that was really sort of the vision that I think Alisa really pitched to us when she first came to us with this project.

I want to ask about this idea of a battle of imaginations, and that's your project overview, your original statement about what you're contending with here. Oh, look, a lot has changed in this subject in particular since 2024 with the new administration, with changes that have been made to federal funding for broadband, and a lot of contentious events that have occurred. What is this battle of imaginations, how does that animate all of that activity over the last couple of years?

Yeah, I would say that a lot of times when we're thinking about the federal policymaking space in DC, it continues to operate largely through networks of access and networks of influence and very well resourced interests that end up reinforcing power structures, and those are power structures inevitably of inequality. And we wanted to make sure that we are actually bringing in the voices of consumers into these conversations. I think that a lot of times people can feel comfortable sitting in their ivory towers and telling people what it is that's best for them, but we wanted to hear directly from low income and middle income consumers across the nation.

And we did so by speaking with people in Colorado, speaking with people in New Mexico, speaking with people in Ohio and Georgia. We would love to have spoken to even more people across the nation, but we really thought it was important that when we're having conversations with policymakers, that we're bringing in stories from these consumers because we know that stories can change hearts, we know that data can change minds, and we are really making sure that we want it to be influential in the conversations that we're having about broadband access, broadband affordability, and broadband access.

I want to offer an additional frame that was something that in a discussion that we specifically had early on when we talked about the design of this project in going into communities and having one-on-one focus groups directly with community members. And that's that this project in and of itself was very much sort of like an exercise in democracy. It's like we continue to talk about tech and telecom's policy as if they live or exist independent of sort of day-to-day life. But the reality is we also stopped calling electricity energy policy once we all came to the conclusion that every individual deserves to have the lights on at home. We stopped calling it energy policy, we just started calling it core infrastructure.

And I think the reality is that we haven't really made that decision just yet when it comes to the internet, which I think it's almost kind of baffling that we haven't made that decision yet given where we are with everything related to AI and the out-sized footprint that big tech and AI companies continue to have not only within our economy, but also certainly even within our democracy and our democratic processes.

And so in 2026, the other thing we really wanted to do with this project is that by 2026, really assert and underscore this notion that internet access is itself core infrastructure in the very same way that electricity is, that running water is, that roads are, which I also might say it's not a coincidence that these are also all the issues that tie into the data center issues that we're seeing right now as well is what happens to utilities writ large for consumers when something like a data center enters our neighborhoods.

And so I think it's interesting, that's a very concrete way of demonstrating the fact that broadband is itself its own connective tissue. The internet is a connective tissue that runs through every single issue area that our families face, whether it's healthcare access, whether it's education, employment, job mobility, even civic participation or housing. A lot of these really depend on your ability to access the internet and actually have the skills, the appropriate skills, to actually engage on these issues online.

And so I think that the other thing that happened at the same time is that this project coincided with the winding down of the ACP program. While we were doing these focus groups, it also coincided with the unlawful cancellation of the Digital Equity Act programming. And so it also was kind of an exercise in democracy in the sense that we were in live time able to capture some of the responses or some of the impacts of those programs ending, of those programs winding down, and actually capture the voices and the attitudes of what people most impacted by them, what they thought about the fact that those programs were wound down and if those programs could come back or if new programs could take their place, what those would look like and how they should be structured.

And so yeah, I always like to really uplift at the very top of our conversation about this paper about the fact that this was also such an exercise in seeing how our democratic processes need to evolve in the wake of the out-sized world that the internet, digital platforms, and increasingly AI-driven economy, what that looks like and how we need to modernize our own democratic processes to reflect this sort of new world that we're living in.

And if I could just quickly add, another quote that we had in our report was by Eddie Glaude Jr. He said, "We've got to struggle democratically if we're struggling for democratic ends, and you've got to lead democratically if you want democratic ends," so we were working to embody that.

And then I would also just add on the timeline piece that I didn't get to mention earlier. I know that Claudia and I were on a panel back in February at the National Digital Inclusion Alliance Conference, Net Inclusion, which took place in Chicago. And I know I made a joke that my friend Joey from Shelby may not have liked, but I said it's like a Rihanna album, it takes time. We just always kept saying, "It's coming out soon. It's coming out soon." But we were able to get this over the finish line in May of this year. And I actually think to Claudia's point, that it actually made it better because if we would have put this project out summer of last year, there's so many changes that have happened since then. So I think that this is the right time, especially given all the conversations going on right now about universal service fund reform.

I think one of the things that I just sort of intuited about these issues is that the most frustrating thing is that many of the solutions were on the table, we appear to be moving in the right direction, not perfectly, not without obstacles or difficulties still ahead, etc., but enormous amount of work, which has essentially been somewhat scrapped in the current policy moment. But I kind of get one of the headline findings of your work here is that this is about political will on some level. The solutions are there. I'm interested in maybe a couple of other examples where some of those focus group participants, some of the kitchen table conversations that you had, where their diagnosis was sharper than the policy framing you hear there in Washington or really cut to the quick of this.

I would start with something that I think working at Unidos has always been very obvious to us, or to me at least, given the sort of unique needs of Latino communities, and that's actually the piece around cultural competency and linguistic competency. So the greatest thing I think that we uplifted or we were able to surface out of these focus groups is the notion that I think what we heard really kind of doubled down on this idea that the Digital Equity Act itself was truly revolutionary. And it was revolutionary in the sense that it gave a lot of agency and ownership to local regions, to states, to communities to actually have say an agency over the ways in which a federal program would fit and be molded into their neighborhood, into their local community, and what actually needed to be in place in order for that to be successful.

So one of the things that we heard directly from participants, let's say for example out of New Mexico and Albuquerque, was this idea that any kind of programs that were going to be delivered from the federal government needed to be done in ways that were in language, that were actually done with an assessment, let's say, of what the community needs. So for example, another focus group participant mentioned how they lived in a rural area and that presented unique challenges from even from some of their friends or family members that might've lived in more urban areas. So it really was the fact that even within small neighborhood or communities, there could be such disparate needs that those needed to be addressed first and foremost. So I think that was something that we heard a lot of.

I think related, and I think this is also kind of another example of something not necessarily being culturally competent but it kind of showed up in a different way, was one of our focus group participants out of our Appalachia interviews in Ohio, in a very rural area in Ohio, mentioned how another issue that they had is that whenever they tried to do broadband access and sign up tables or tents at local farmer's markets or at local markets, another issue they ran into was the fact that a lot of their community members themselves were not literate, that they straight up cannot read and write at even a basic level. And so that even if you have this wonderful program that gives away, that provides a benefit, that provides a discounted or free device that someone can use, even if you are able to deliver a one-on-one digital training course, if you are not realizing that your community can't even enter that barrier, that they're already facing a barrier of entry that they cannot surpass, that that was never going to be a viable solution for that community.

And so I think again, coming back to this idea that the program of the Digital Equity Act was really revolutionary in the sense that it really provided a lot of that self-direction that we really see is missing from any sort of the other legislative proposals that we even see pending right now. And so again, that's why I think the unlawful cancellation of the program is extremely disappointing because of how many steps forward it took us in terms of actually addressing the very needs of the community members who need it most.

Latest

Podcast

Recent

Our content delivered to your inbox

Thank you!

Around the globe