Cohere and Aleph Alpha Merge Into a $20 Billion Rival

Two national AI champions combine rather than compete

Cohere and Aleph Alpha signed a definitive agreement on September 16 to merge, converting a plan first floated in April into a binding deal. The combined company carries a valuation of roughly $20 billion, keeps the Cohere brand and Aidan Gomez at the helm, and runs from dual headquarters in Toronto and Berlin, with Aleph Alpha's Heidelberg base becoming a dedicated research center.

The financial spine of the deal is enterprise revenue and patient capital rather than consumer scale. Cohere was valued at $7 billion in September 2025 and reported about $240 million in revenue for the year, built on its Command models, the Parse document engine, and the North productivity assistant. Aleph Alpha brings a decade of work on smaller, controllable models tuned for European governments and regulated firms, along with research on tracing why a model produced a given answer. Germany's Schwarz Group, the retail conglomerate behind Lidl and the STACKIT cloud, is committing 500 million euros and deepening an infrastructure partnership, and The Globe and Mail has reported a Canadian government-backed consortium weighing an investment of around $3 billion. The message the two companies lead with is explicit. They call the result the first transatlantic sovereign AI provider.

Why sovereignty became the defensible position for mid-tier labs

The enterprise middle of the AI market has been squeezed from both directions for the past year. Above it sit OpenAI, Anthropic, and Google, spending tens of billions on training runs that a $240 million revenue base cannot begin to match. Below it sit Chinese open-weight models that keep closing the capability gap and drag the price of a raw API call toward zero. A lab caught between those two forces cannot win on frontier benchmarks and cannot win on price. It has to win on something the giants structurally cannot offer.

Sovereignty is that something. A German bank, a Canadian ministry, or a European hospital often cannot send its data to a US frontier lab under a foreign cloud's control, regardless of how good the model is. Data residency, national compute, and auditable deployment inside a customer's own borders form a moat that OpenAI and Google cannot easily cross, because their scale is tied to jurisdictions and business models that regulated buyers are trying to avoid. By fusing a Canadian enterprise vendor with a German government supplier and anchoring both to European cloud infrastructure, the merger builds a company whose entire pitch is the one thing the frontier labs cannot credibly sell.

The compute and regulatory tailwind behind the deal

The timing also tracks a regulatory environment that keeps tilting toward exactly this kind of company. The European Union's AI Act is now in active enforcement, sovereign cloud mandates are spreading through national procurement rules, and public bodies across Europe and Canada increasingly treat data residency as a hard requirement rather than a preference. Every one of those pressures raises the value of a vendor that can promise a model trained, hosted, and audited without leaving a friendly jurisdiction. Aleph Alpha spent years courting German ministries and defense-adjacent buyers on precisely that promise, and Cohere brings the enterprise sales motion and model quality to scale it across the Atlantic.

Compute is the other half of the equation, and it is the part most coverage understates. Neither company can rent frontier-scale capacity from a US hyperscaler without reintroducing the dependency its customers are paying to avoid. The Schwarz Group relationship and its STACKIT cloud give the merged firm European infrastructure it can point customers to, and the reported Canadian consortium would add sovereign compute on the other side of the ocean. The deal is as much about securing places to train and serve models as it is about combining two model families. It echoes the logic behind Mistral's own push toward European sovereignty, where infrastructure and national backing matter as much as the model.

What $240 million in revenue says about the enterprise LLM market

The gap between a $20 billion valuation and $240 million in revenue is the real story inside the numbers. It shows how far enterprise AI still sits from the recurring revenue that would justify these valuations on fundamentals, and it explains why consolidation, not competition, is becoming the default move for everyone outside the top three. Merging buys three things that are hard to build alone at this stage: distribution into two continents of government and enterprise accounts, a balance sheet deep enough to fund the next model, and guaranteed compute through the Schwarz and STACKIT relationship. For a standalone lab burning cash toward an uncertain frontier, acquiring those through a partner is cheaper than earning them one contract at a time.

That logic points at the dozen other mid-tier labs watching this deal. The independent enterprise model vendors that raised at high valuations in 2024 and 2025 face the same math Cohere and Aleph Alpha just resolved, and most of them lack a Schwarz Group willing to write a nine-figure check. Sovereignty gives the survivors a reason to exist that scale and price never could, and this merger turns that reason into a corporate structure other regional champions will study closely. The AI market's consolidation has begun at the middle, and the labs that endure there will be the ones that sell trust and jurisdiction rather than raw capability.