TL;DR
German AI company Aleph Alpha announced a combination with Canada’s Cohere, creating a business valued at about $20 billion with headquarters in Toronto and Heidelberg. The deal strengthens its financial position but leaves open how much control Germany retains over the models underpinning its sovereign AI plans.
Germany’s Aleph Alpha announced a combination with Canadian AI developer Cohere on April 24, creating a company valued at about $20 billion as Germany invests heavily in sovereign AI infrastructure. The transaction is not confirmed as an outright sale, but it places the future direction of Germany’s best-known independent model developer under a cross-border structure.
The combined company will maintain dual headquarters in Toronto and Heidelberg, according to the supplied deal information. Germany’s Schwarz Group, the owner of Lidl, Kaufland and cloud provider StackIT, is leading a $600 million investment in Cohere’s Series E financing. The companies plan to offer their technology through StackIT.
The combination gives Aleph Alpha access to a larger balance sheet and a broader commercial operation. Cohere has positioned private deployments and sovereign AI as core products, while Aleph Alpha had struggled to match the resources and market reach of OpenAI, Google and other global developers. Heidelberg remains part of the corporate structure, but future decisions about model development will also involve Toronto.
The deal arrives as Germany expands the infrastructure beneath its AI market. Deutsche Telekom and NVIDIA opened an Industrial AI Cloud in Munich on February 4 with almost 10,000 Blackwell GPUs and about 0.5 ExaFLOPS of computing capacity. Telekom says the privately financed system increases Germany’s AI computing capacity by roughly 50%, with SAP serving as a platform partner.
Der Souveränitäts-Markt ist real geworden —
und hat im selben Quartal seinen Champion verkauft
Tagesaktuell verifizierter Marktpuls · Geld, GPUs und eine Ironie
Das Geld ist da — drei Belege
Telekom + NVIDIA in München: ~0,5 ExaFLOPS, +50 % deutsche KI-Rechenleistung, privat finanziert. Schwarz-Gruppe: 11 Mrd. €, perspektivisch 100.000 GPUs.
805 Mio. € Gigafactory-Förderung; Konsortium SAP, Telekom, Siemens, IONOS, Schwarz. SPRIND: 125 Mio. € für eigene KI-Labore.
BfV wählt ChapsVision statt Palantir; Bundeswehr schließt Palantir aus der Cloud aus. Gartner: EU-Sovereign-Cloud +83 % auf 12,6 Mrd. $.
DIE IRONIE · 24. APRIL 2026
Mitten im Souveränitäts-Frühling schließt sich Aleph Alpha mit Kanadas Cohere zusammen — die Schwarz-Gruppe finanziert als Lead-Investor mit 600 Mio. $.
Freundliche Lesart: Konsolidierung unter Gleichgesinnten; 20 Mrd. $ Verbund schlägt unterfinanziertes Startup. Unbequeme Lesart: Deutschlands Modellschicht wird künftig in Toronto mitentschieden — und deutsches Kapital finanziert lieber fremde Champions als eigene.
Souveränität ist eine Schichtenfrage
Das Signal: Die souveräne Betriebsschicht ist jetzt kaufbar und bezahlbar — die Modellschicht bleibt Import. Wer Souveränitätsstrategien baut, sollte sie auf die Schichten bauen, die Europa tatsächlich kontrolliert.
ZIMA CUBE Creator Pack AI NAS & Local AI Server
RTX PRO 2000 GPU for Local AI & Creator Workflows: Built for GPU-accelerated local AI inference, AI agents,…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Control Shifts Above the Data Center
The combination exposes a gap in Europe’s sovereignty strategy. Germany can place servers, data and system access under domestic operators and German law, yet the model layer may still depend on companies whose governance, research priorities or ownership extend beyond Europe.
Supporters can view the transaction as cross-border consolidation between companies that both sell private AI systems. A better-funded business may compete more effectively for government and corporate contracts. The opposing reading is that German capital is financing a multinational champion after Aleph Alpha failed to become a fully independent national model provider. Both interpretations depend on governance details that have not been disclosed.
FDE: The Forward Deployed Engineer: Architecting the Last Mile of Enterprise AI
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Germany Builds Its Computing Base
Public and private spending shows that sovereign AI has moved into procurement and infrastructure. German budget documents allocate €805 million in 2026 toward attracting a European AI gigafactory. A consortium involving SAP, Deutsche Telekom, Siemens, IONOS and Schwarz Group has been discussing a joint European Union bid.
Schwarz Group is also reported to be planning about €11 billion in cloud and AI investment, potentially supporting 100,000 GPUs, while the federal innovation agency SPRIND has launched Next Frontier AI with €125 million for AI laboratories. The precise schedules and final capacity of these projects have not been publicly established in the supplied material.
Demand is also appearing in government purchasing. Germany’s domestic intelligence agency selected France’s ChapsVision rather than Palantir in May, while the German armed forces excluded Palantir from cloud projects. Gartner forecasts European sovereign-cloud spending of $12.6 billion in 2026, an 83% annual increase. McKinsey separately estimates that sovereign AI could represent nearly $600 billion of a global AI services market exceeding $1 trillion, though that figure remains a consultancy projection rather than recorded revenue.
“The combined company will maintain dual headquarters in Toronto and Heidelberg and make its joint offering available through StackIT.”
— Cohere and Aleph Alpha deal announcement
sovereign AI cloud solutions
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Governance Terms Remain Undisclosed
It is not yet clear how ownership, voting rights and board control will be divided, where the combined company’s main research decisions will be made or which legal jurisdictions will govern particular models and customer data. The supplied information also does not establish whether Aleph Alpha was legally acquired, merged on equal terms or placed under another corporate arrangement. Calling the company sold goes beyond the confirmed facts available.
Germany’s hardware independence is also limited. The new facilities rely heavily on NVIDIA processors, leaving an important part of the supply chain tied to a US-based vendor even when the computers and data remain in Germany.
ENTERPRISE AI INFRASTRUCTURE: Modern MLOps, Vector Databases, GPU Clusters, and Scalable Data Architecture for LLMs (The Enterprise AI Architect’s Handbook)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Deal Terms and EU Funding Loom
The next markers will be publication of the final transaction structure, regulatory reviews and details of how research, intellectual property and customer contracts are divided between Toronto and Heidelberg. Governments and corporate buyers will also test whether the joint Cohere-Aleph Alpha offering meets their data residency and operational-control requirements. Germany’s gigafactory bid and the rollout of StackIT capacity will show whether domestic infrastructure spending produces sustained demand.
Key Questions
Was Aleph Alpha sold to Cohere?
The companies announced a combination, but the supplied information does not disclose enough ownership or voting detail to confirm an outright sale.
Will Aleph Alpha leave Germany?
No departure has been announced. The combined company is expected to retain Heidelberg alongside a second headquarters in Toronto.
Who is financing the deal?
Schwarz Group is leading a $600 million investment in Cohere’s Series E financing and plans to support the joint offering through StackIT.
Does Germany now have sovereign AI infrastructure?
Germany has expanded domestically operated computing capacity, including the Munich cloud. Full independence is not established because models and NVIDIA hardware still involve foreign suppliers.
Why does the model layer matter?
The model developer controls choices involving training, updates and technical access. Domestic data centers alone do not give a country complete authority over those model-level decisions.
Source: Thorsten Meyer AI
