TORONTO / HEIDELBERG, 5 MAY 2026 — Cohere, the Canadian enterprise AI company last valued at US$6.8 billion, and Aleph Alpha, Germany's most prominent AI laboratory, have merged to create what both companies and their respective government backers are describing as a "sovereign AI" alternative to the United States-China duopoly that currently controls global frontier artificial intelligence development. The merger, blessed by both the Canadian and German governments, is the most significant consolidation in European and Canadian AI to date — and its strategic ambitions are significantly more complex to achieve than the branding implies.

The merger arrives in a week when the geopolitical dimensions of AI have become impossible to ignore. On the same day Cohere and Aleph Alpha announced their combination, China's National Development and Reform Commission issued its first-ever prohibition of a foreign acquisition of a Chinese AI company, blocking Meta's reported US$2 billion offer for Manus, a Chinese AI agent startup. The symmetry is pointed: one side of the duopoly is blocking inbound acquisition; the other is arguing that a merged European-Canadian entity can provide an alternative to both. As Nathan Benaich observes in his State of AI report for May 2026, "the strategic substance behind the framing has not yet been tested."

Who Cohere and Aleph Alpha Are

Cohere was founded in 2019 by Aidan Gomez — a co-author of the seminal "Attention Is All You Need" transformer paper published at Google in 2017 — alongside Ivan Zhang and Nick Frosst. The company built its business around enterprise language model deployment, with specific strengths in retrieval-augmented generation, text classification, and the Command R model family designed explicitly for business use cases rather than consumer chat. Cohere's customer base is concentrated in financial services, healthcare, and government, and the company's deployment flexibility — models available on Cohere's cloud, on AWS, Azure, GCP, or on-premises — made it attractive to enterprises with data residency requirements that ruled out US-only cloud deployments.

Aleph Alpha was founded in 2019 in Heidelberg, Germany by Jonas Andrulis and Samuel Weinbach with an explicit focus on European sovereignty and regulatory compliance. The company's Luminous model family was designed to meet European data protection requirements by default, with deployment options that kept data within European Union jurisdictional boundaries. Aleph Alpha has worked closely with the German federal government, the European Defence Agency, and several EU member state governments on AI deployments in sensitive contexts — including defence, intelligence, and public administration — where American or Chinese cloud providers are considered unsuitable on national security grounds.

The combined entity brings together Cohere's enterprise commercial traction and infrastructure scale with Aleph Alpha's European government relationships and regulatory pedigree. On paper, the strategic complementarity is genuine: Cohere provides the commercial engine and the model capability; Aleph Alpha provides the institutional relationships and the regulatory credibility that European government contracts require. The merged company would be one of the few AI vendors globally capable of deploying enterprise-grade language models in settings where neither US nor Chinese providers are acceptable counterparties.

Why Governments Endorsed It

Canadian Prime Minister Justin Trudeau's government and the German federal government under Chancellor Olaf Scholz both publicly endorsed the merger. The endorsements are not incidental — they reflect a strategic calculus that both governments have been developing since 2023, when the concentration of frontier AI capability in OpenAI, Google, Anthropic, and a small set of Chinese labs became impossible to characterise as anything other than a geopolitical dependency risk.

For Canada, Cohere represents one of the country's most significant commercial AI assets, built by a team that emerged from the University of Toronto's world-leading machine learning research community. The Canadian government's concern is straightforward: if Cohere were acquired by a US or Chinese company, the intellectual property, the talent, and the institutional relationships would migrate with it, leaving Canada without a domestic AI champion capable of serving its own government and regulated industries on sovereignty terms.

For Germany, Aleph Alpha has served as a proof-of-concept for European AI capability and a vehicle for German government AI deployments that require European data residency and European legal jurisdiction. The German government's enthusiasm for the merger reflects its assessment that Aleph Alpha alone is insufficiently capitalised to reach the scale required to compete for large enterprise and government contracts against OpenAI or Google, while a merged entity with Cohere's resources and commercial track record might be.

The Strategic Gap: What the Merged Company Still Lacks

Government endorsement and strategic complementarity do not, by themselves, close the capability gap between Cohere-Aleph Alpha and the frontier labs. On standard benchmark comparisons, Cohere's Command R+ and Aleph Alpha's Luminous models perform creditably in their target domains — enterprise retrieval and European-language tasks respectively — but both lag materially behind GPT-4o, Claude Opus 4.7, and Gemini 1.5 Pro on general-capability evaluations. The gap is not static, and the merged entity will have more resources to close it, but at the time of the merger announcement, it is real.

The more fundamental challenge is compute. Training frontier AI models at the scale that OpenAI, Google, and Anthropic operate requires access to tens of thousands of the most advanced GPU clusters — hardware that is predominantly manufactured by US companies (NVIDIA, AMD) under US export control regimes, and that requires massive data centre investments that few non-hyperscale companies can sustain. The merged Cohere-Aleph Alpha entity is better capitalised than either company alone, but it is not operating at the compute scale of the frontier labs, and closing that gap requires capital investment that government endorsement alone does not provide.

Benaich's assessment — that the strategic substance "has not yet been tested" — reflects the gap between the framing and the reality. "Sovereign AI" is a compelling political narrative, but political narratives do not run inference. The question for enterprise and government customers evaluating the merged entity is whether it can deliver capabilities at the level their use cases require, on the deployment terms their compliance requirements mandate, at competitive pricing. Those questions will be answered by the product roadmap over the next 12 to 18 months, not by the merger announcement.

China's Manus Veto: The Other Side of AI Geopolitics

The National Development and Reform Commission's prohibition of Meta's acquisition of Manus represents a structural shift in how China manages its AI industry — and a signal that is relevant for any company or government thinking about AI vendor relationships with Chinese labs. Manus, the Chinese AI agent startup that Meta was reportedly willing to pay US$2 billion to acquire, builds autonomous AI agents capable of executing complex multi-step tasks across software environments. The Chinese government's decision to block the acquisition on national security grounds treats Manus not as a commercial asset to be sold to the highest bidder but as a strategic technology that must remain under Chinese ownership and control.

The veto establishes a precedent: Chinese AI companies with capabilities deemed strategically significant are not available for foreign acquisition regardless of price. This principle, applied consistently, means that the open-source releases from DeepSeek, Moonshot, MiniMax, and Z.ai — which have been providing Western and ASEAN developers with access to near-frontier Chinese AI models — do not translate into the ability to acquire the companies or the teams that built them. China's AI is available to use but not to own.

ASEAN's Sovereignty Question

For Southeast Asian governments, the emergence of Cohere-Aleph Alpha as a claimed sovereign AI alternative raises a set of questions that are simultaneously strategic, economic, and practical. Singapore's National AI Strategy 2.0, published by IMDA in 2024, explicitly discusses reducing dependence on any single AI vendor while building domestic AI capabilities. Indonesia, under President Prabowo Subianto, has articulated ambitions for Indonesian-language AI models that reflect and reinforce the country's digital sovereignty priorities. Malaysia has positioned itself as a regional AI hub, signing infrastructure and capability agreements with Microsoft, Google, and NVIDIA while developing domestic AI training programmes. Vietnam's Vingroup has invested in VinBigdata, which has been developing Vietnamese-language models.

The economics of full AI sovereignty — developing and maintaining frontier-capable models independently — are prohibitive for all but the wealthiest states. The compute costs alone for training a GPT-4-class model are estimated at US$100 million or more, and the engineering talent required is scarce globally. What ASEAN governments can realistically pursue is a "distributed sovereignty" model: using open-source models that can be self-hosted within national jurisdictions (DeepSeek, Llama 3, Mistral) for sensitive applications, while maintaining relationships with multiple Western and potentially European AI providers for enterprise deployments. The Cohere-Aleph Alpha merger adds a third option to that portfolio — one that explicitly offers European regulatory compliance and government-endorsed deployment models.

Singapore's approach to AI sovereignty is likely to remain characteristically pragmatic. The city-state's interests are served by maintaining relationships with multiple AI providers — US, European, and where appropriate open-source Chinese models — rather than committing to any single vendor's ecosystem. The Cohere-Aleph Alpha merger creates a more credible European option for Singapore enterprises and government agencies with specific data residency or regulatory requirements, but it does not change the fundamental calculus: Singapore will continue optimising for capability, compliance, and competition rather than ideological alignment.

Sources

  • Nathan Benaich — State of AI, May 2026
  • Cohere — Merger Announcement, May 2026
  • Aleph Alpha — Merger Announcement, May 2026
  • China National Development and Reform Commission — Manus Acquisition Statement, 2026
  • IMDA Singapore — National AI Strategy 2.0, 2024
  • European Commission — AI Act Implementation Guidelines, 2025