Cohere just bet $20 billion on sovereignty

On April 24, Cohere announced it would merge with Germany's Aleph Alpha in a deal valuing the combined entity at $20 billion. Schwarz Group, the parent company of Lidl and one of Aleph Alpha's top backers, is investing $600 million to lead Cohere's Series E. Both the Canadian and German governments helped facilitate the deal. Cohere's shareholders will hold roughly 90% of the new company, with Aleph Alpha taking the remaining 10%. This is not a product announcement. It is a geopolitical positioning move. Cohere is betting that sovereignty, the ability to deploy, govern, and control AI without depending on American hyperscalers, is about to become the defining purchase criterion for European enterprises and governments. If they are right, this merger could create the first credible non-US, non-Chinese enterprise AI platform. If they are wrong, they just overpaid for a pivot.

The setup

The timing is not accidental. Europe is caught in a squeeze. On one side, the US State Department is warning allies about Chinese AI espionage and pushing for tighter technology controls. On the other, American companies are hoovering up European enterprise data through cloud services governed by US law. The CLOUD Act gives US law enforcement the power to compel American companies to hand over data stored abroad, even if the servers sit in Frankfurt or Amsterdam. Selecting "EU region" in AWS or Azure does not guarantee sovereignty if the provider is US-headquartered. The legal jurisdiction follows the company, not the data center. GDPR already made data residency a real constraint. Meta paid €1.2 billion for transferring EU user data to the United States. TikTok followed with €530 million. Uber paid €290 million. These are not theoretical risks. They are line items on balance sheets. According to Deloitte, 73% of enterprises now cite data privacy and security as their top AI risk concern, and 77% factor a vendor's country of origin into AI purchasing decisions. Cohere is betting that "AI sovereignty" is the enterprise version of GDPR compliance, not optional, but table stakes. The pitch is simple: you should not have to send your most sensitive data to an American cloud to use AI.

What Cohere actually gets

Aleph Alpha is not a frontier model lab. It pivoted away from building large language models and toward enterprise AI applications. What it does have is something Cohere cannot easily build: deep institutional relationships with the German public sector and regulated European industries. Those government pipelines are the real asset. Cohere, for its part, brings global AI scale. Founded in 2019 by former Google researchers including Aidan Gomez (a co-author of the original Transformer paper), the company has built a reputation for enterprise-grade AI that prioritizes deployability and data privacy. It already raised $1.6 billion from investors including Nvidia and AMD before this deal. The merger gives Cohere an anchor in Europe's largest economy and a direct line into the institutions most likely to mandate sovereign AI solutions. The combined entity positions itself as what the press release calls "a transatlantic AI powerhouse, anchored in Germany and Canada." The framing matters. This is not an American company opening a European office. It is a Canadian-German alliance explicitly designed as a counterweight to Silicon Valley.

Why cloud sovereignty failed but AI sovereignty might not

Europe has tried this before. The cloud sovereignty wave of the 2010s produced companies like OVHcloud, Scaleway, and Hetzner, European-owned alternatives to AWS and Azure. The results have been sobering. European cloud providers make up just 15% of their own market, down from 29% in 2017. Synergy Research called the hyperscalers' dominance "an impossible hill to climb." The EU Commission is now considering ditching Microsoft Azure for OVHcloud for its own infrastructure, but that is more a political statement than a market shift. The cloud providers failed because infrastructure is a scale game. AWS, Azure, and Google Cloud can outspend any European competitor on data centers, networking, and developer tooling. The service catalog gap is real. Scaleway is 60% cheaper on some workloads, but it offers fewer advanced features. For most enterprises, the convenience of the hyperscaler ecosystem outweighed the sovereignty argument. AI might be different, and the reason is stickiness. Cloud infrastructure is relatively fungible. You can migrate a Kubernetes cluster from AWS to Scaleway with effort but without rethinking your entire stack. AI models are not like that. Once an enterprise fine-tunes a model on proprietary data, builds workflows around its behavior, and integrates it into decision-making processes, switching costs become enormous. The model layer creates lock-in that the infrastructure layer never could. Cohere is betting on this asymmetry. If they can get European enterprises to build on Cohere's models and deploy them on sovereign infrastructure, the switching costs will keep those customers locked in for years. The model layer, not the cloud layer, is where sovereignty actually sticks.

The American version of the same play

Cohere is not the only company betting on vertical integration as sovereignty. Elon Musk's Terafab project, announced in March 2026, is the American mirror image. Terafab is a planned semiconductor fabrication complex in Austin, Texas, jointly developed by Tesla, SpaceX, and xAI. The facility aims to consolidate chip design, fabrication, memory production, advanced packaging, and testing under one roof, producing more than one terawatt of AI compute capacity per year. Tesla plans to use Intel's 14A manufacturing process, which would mark Intel's first major customer for that technology. The logic is identical to Cohere's, just applied at a different layer of the stack. Musk does not want his AI ambitions dependent on TSMC's fabrication capacity in Taiwan or Nvidia's GPU allocation decisions. He wants to own the full pipeline. It is vertical integration as sovereignty, except the threat Musk is hedging against is not regulatory. It is supply chain fragility. Both moves reflect the same underlying thesis: in an era where AI is becoming critical infrastructure, depending on someone else's stack is an unacceptable strategic risk. Cohere is selling that thesis to European governments. Musk is building it for himself.

The middle powers are moving

Cohere's merger is part of a broader pattern. A new geopolitical category is emerging in AI: the middle powers. Countries that are not the US or China but refuse to simply rent their AI capabilities from either. Japan's Sakana AI, founded by former Google researchers, is building domestic AI infrastructure. South Korea is backing a massive state-led AI ecosystem. Israel's AI21 Labs is reportedly exploring strategic options. The UAE's G42 is investing heavily in AI compute. In Europe, Germany has DeepL and Black Forest Labs. France has Mistral. The Netherlands has Axelera AI. What these efforts share is a recognition that AI capability is becoming a prerequisite for economic competitiveness, and that outsourcing it entirely to the US or China creates dependencies that are hard to reverse. Carnegie's Anton Leicht put it directly: the AI middle powers "need to find a strategy for participating in rapid AI progress," and their participation depends on finding a niche, securing access to compute, and building defensive capabilities. Cohere's niche is clear: be the sovereign enterprise AI provider for organizations that cannot or will not depend on American tech. The question is whether that niche is big enough to sustain a $20 billion valuation.

The incumbent problem

Here is where the thesis gets harder to defend. Google, Microsoft, and Amazon are not ignoring the sovereignty demand. They are co-opting it. AWS launched its European Sovereign Cloud in January 2026, an isolated cloud environment entirely within the EU, governed by a European board and staffed by European employees. Microsoft offers Azure sovereign cloud configurations with encryption keys managed locally. OpenAI now offers data residency in Europe for ChatGPT Enterprise and its API platform. Google expanded its AI investments in Singapore and is running sovereign cloud deployments for governments across Asia. These "quasi-sovereign" solutions are not true sovereignty, as critics correctly point out. The US CLOUD Act still applies because the parent company is American. But for many enterprises, the mitigation is thick enough. They get EU data residency, local key management, and contractual protections, all on a platform with a vastly larger feature set than anything Cohere or Aleph Alpha can offer today. The Council on Foreign Relations noted that hyperscaler sovereign clouds "may be the closest our allies come to true digital sovereignty." That is both a concession and a competitive threat. If 80% sovereignty on AWS is good enough for 80% of enterprises, Cohere's addressable market shrinks to the remaining 20%, the most regulated, most sensitive, most government-adjacent workloads. That is still a large market. But it is a different business than building a general-purpose AI platform.

Regional models as a wedge

One angle that could expand Cohere's market is the rise of regional AI models. The logic goes like this: a model fine-tuned on French legal documents, German industrial data, or Singaporean regulatory frameworks will outperform a generic global model on those specific tasks. If Cohere can build a portfolio of regionally specialized models, each deployed on sovereign infrastructure, it creates value that the hyperscalers cannot easily replicate. This is not hypothetical. The Linux Foundation found that 82% of organizations are already developing customized AI solutions to maintain control. The demand for models that understand local languages, cultural context, and regulatory nuance is real. Aleph Alpha's existing relationships with German institutions give Cohere a head start in exactly this kind of localization. The risk is that open-weight models make this advantage temporary. Meta's Llama, Alibaba's Qwen, and Mistral's models are all freely available for fine-tuning. Any enterprise can download an open model, fine-tune it on local data, and deploy it on whatever infrastructure they choose. Cohere's edge has to be more than just "we offer a model you can run in Europe." It has to be a full-stack sovereign AI experience that is easier, faster, and more reliable than the DIY alternative.

What this actually means

Cohere's $20 billion bet is a wager on a specific future: one where the geopolitical fragmentation of AI accelerates, where data sovereignty requirements tighten, and where enterprises increasingly demand AI solutions that operate entirely outside American legal jurisdiction. If that future arrives, Cohere is extraordinarily well-positioned. A transatlantic entity anchored in two G7 nations, with government backing on both sides, deep enterprise relationships in Europe's largest economy, and a product built from the ground up for sovereign deployment. That is a combination no one else has. But the future is not guaranteed. If hyperscaler sovereign clouds prove "good enough" for most enterprises, if open-weight models commoditize the AI layer, or if the geopolitical winds shift toward cooperation rather than fragmentation, then Cohere may have overpaid for a thesis that does not fully materialize. The honest assessment is that sovereignty is becoming a real purchasing criterion for a meaningful segment of the market, but not the entire market. Cohere does not need to win everywhere. It needs to win in the places where sovereignty is non-negotiable: defense, healthcare, public sector, critical infrastructure, and heavily regulated industries. That is a narrower market than "enterprise AI," but it is a deep one, and it is growing. The Cohere-Aleph Alpha merger is not really about AI models. It is about who gets to control the infrastructure layer of the next economy. That is a $20 billion question, and the answer depends less on technology than on politics.

References