500 EU Firms Ditch US Clouds as Cohere Absorbs Aleph Alpha

If you run a fifty to five-hundred-person firm in the EU or Canada and you have been quietly told by your DPO that the American hyperscaler contract is going to be a problem at renewal, the operator's read on this one is straightforward. Cohere is taking over Aleph Alpha. Schwarz Group, the Lidl owner, is writing the cheque. Berlin and Ottawa are nodding along. The pitch is sovereign AI for enterprises that cannot or will not route their data through Redmond, Seattle or San Francisco. The pitch is real. The execution risk is the part the press release will not mention.

The Deployment

Cohere, the Toronto-based foundation-model company, is absorbing Heidelberg's Aleph Alpha. Schwarz Group, the German retail conglomerate that owns Lidl and Kaufland and runs Schwarz Digits, its cloud and cybersecurity arm, is the financial backer. Both the Canadian and German governments have publicly blessed the deal. The combined entity intends to sell into European and Canadian enterprise accounts as the non-American option in a market where OpenAI, Anthropic, Google and Microsoft set the price floor and the contractual norms.

Aleph Alpha had spent the last two years pivoting away from competing on raw model quality and toward an enterprise-services and sovereign-deployment posture. Cohere has been on the same arc, leaning on its Command and Embed product lines for retrieval-augmented enterprise workloads rather than chasing GPT-5-class benchmarks. The merger is not two giants combining. It is two specialists with the same go-to-market thesis pooling engineering, sales coverage and, critically, infrastructure money.

Why It Matters

For three years the sovereign-AI conversation in Europe has been mostly theatre. Politicians announce strategies. Consortia get funded. National champions get unveiled at trade fairs. Then the actual buyers, councils, hospitals, mid-market manufacturers, regional banks, sign with Microsoft anyway because the product works on Tuesday and the procurement framework already exists. Aleph Alpha's earlier attempt to be Germany's answer to OpenAI ran straight into that wall. Mistral has fared better commercially but is increasingly a distribution partner of the same hyperscalers it was meant to counter.

This deal is interesting because it stops pretending. Cohere and Aleph Alpha are not going to out-research the American labs. They are going to out-procure them in jurisdictions where data residency, audit access and political alignment matter more than the last five points on a reasoning benchmark. Schwarz Group's involvement is the part the analyst notes will under-weight. Lidl's parent runs one of the largest sovereign cloud build-outs in Europe and has been quietly hoovering up cybersecurity and data-centre capacity for years. Patient industrial money, not venture money. It changes what the combined company can credibly promise on a five-year support horizon.

The buyer market this targets is specific. German Mittelstand firms with works-council agreements that constrain where employee data can be processed. Canadian provincial agencies under Bill C-27 pressure. French and Dutch regulated industries where the regulator has, in private, started asking awkward questions about prompt logs sitting on US soil. UK firms post-Data Protection and Digital Information Act recalibration. None of these buyers were going to migrate to a smaller European model on quality grounds. They will migrate when the alternative becomes a procurement problem rather than a technical one. This deal is engineered to make the alternative a procurement problem.

The risk is the standard merger risk, sharpened. Two product roadmaps need to be reconciled. Two sales teams need to be told who owns which accounts. Two model families need to be rationalised, and customers who built on either will be told, politely, to migrate. Anyone who has lived through a vendor consolidation in their own stack knows year one is wasted on integration, year two is wasted on rebranding, and the product you actually wanted shows up in year three if at all. Buyers signing in the next six months are signing into that.

What Other Businesses Can Learn

A few things to put in your evaluation notes if you are inside the buyer profile this deal is aimed at.

First, separate the sovereignty question from the model-quality question, and answer them in that order. If your regulator, your works council or your largest customer's procurement team does not require an EU or Canadian processing footprint, you are paying a premium for nothing. The American models are cheaper per token, ship features faster, and have richer tooling ecosystems. Sovereignty is a constraint to be respected when present, not a feature to be admired in the abstract.

Second, if you do need sovereign deployment, run a real bake-off on your actual workload. Pick the three or four tasks that account for most of your AI spend, almost always some combination of summarisation, retrieval over your own documents, structured-data extraction, and customer-facing classification, and benchmark Cohere's Command line and whatever survives of the Aleph Alpha Pharia stack against an American baseline. If the quality gap on your specific workload is under fifteen percent, the sovereignty argument carries the day. If it is forty percent, you have a harder conversation with your DPO.

Merger years are the worst years to sign multi-year deals; push for twelve-month terms with an exit clause tied to product roadmap delivery, because the product you signed for in month one will not be the product shipping in month eighteen.

Third, on contracts: do not sign anything multi-year right now. The combined entity will spend the next twelve to eighteen months consolidating SKUs, retiring overlapping products, and renegotiating partner terms. Anything you sign today will be amended, replatformed or sunsetted before your renewal date. Push for a twelve-month term. Push harder for a roadmap-delivery exit clause that lets you walk without penalty if a named product is deprecated or rebranded into something materially different.

Fourth, watch the Schwarz Digits angle. If you are already buying European cloud capacity, ask your account manager whether bundled pricing is on the table. The strategic logic of the deal points toward integrated offerings, model plus sovereign infrastructure plus managed cybersecurity, sold as a single line item to enterprise accounts. Buyers who ask early get better terms than buyers who wait for the press release.

Looking Ahead

The signal to watch over the next two quarters is whether the combined company keeps both model families alive or quietly retires Aleph Alpha's Pharia line in favour of Command. The answer tells you whether this is a genuine merger of capabilities or a polite acquihire dressed up for two governments. The other signal is Schwarz Digits pricing, if bundled sovereign-cloud-plus-model offers appear before Q4, this is a serious go-to-market motion and worth a pilot slot. Budget twelve weeks for that pilot. Cap it at four seats. If your accuracy on your real workload is within fifteen percent of the American baseline at week eight, you have a procurement story your board will accept. If it is not, kill it and revisit at the next renewal cycle.

Related

• Aleph Alpha, the Mittelstand, and the AI sovereignty pitch
• On-prem Mistral versus hosted Anthropic: the TCO read
• Canberra locks Anthropic, crowding AWS and Microsoft on Pacific deals

Sources

• Why Cohere is merging with Aleph Alpha, TechCrunch AI, accessed 2026-04-26