Sovereign AI  ›  Cohere

Cohere’s Sovereign Gamble

Can Canada’s AI champion deliver strategic autonomy?

Canada has committed to invest up to C$1.6 billion in Cohere as part of a historic financing round. But the real test of sovereignty is whether Canada can preserve its technological choices — even if its national champion fails.

Alaaeddin Al-Jallad · AIGNITRON Inc.Facts current as of 2 Nov 2026Sovereign AI · Part 3

A national champion—and a sovereignty test

Facts current as of November 2, 2026. The Cohere–Aleph Alpha combination remains subject to regulatory approval, the Series E has not completed its final close, and the terms of Canada’s equity participation have not been fully disclosed.

On October 29, 2026, the Government of Canada announced a commitment to make an equity investment of up to C$1.6 billion in Toronto-founded Cohere, marking one of the most consequential commitments yet in Canada’s effort to establish sovereign artificial intelligence capabilities.

This is more than another technology subsidy. Through a commercial equity investment, Canada intends to acquire a stake in a domestic AI company, participate in its potential growth, and secure a stronger position in one of the most strategically important industries of the coming decades.

But the announcement raises a fundamental question that extends well beyond corporate investment or national technology policy:

Can Canada use a private AI company to secure strategic autonomy without becoming dependent on that company itself?

This is the central paradox of Cohere’s position.

Cohere is a Canadian-founded enterprise AI developer with international ambitions, access to global compute infrastructure, and an increasingly important role in the sovereign AI strategies of allied countries. Yet its competitive position depends on technologies, infrastructure, capital, and supply chains that Canada does not independently control.

The challenge, therefore, is not simply to build a successful Canadian AI company. It is to turn that company into an anchor of a broader ecosystem — one that preserves Canada’s ability to develop, deploy, govern, and replace critical AI capabilities regardless of the commercial fortunes of any single provider.

In the opening installment of this series, The Sovereign AI Paradox: Owning the Infrastructure Is Not the Same as Controlling the Technology, I examined why acquiring domestic data centres and graphics processing units (GPUs) does not automatically produce technological independence.

True AI sovereignty, I argued, is not achieved through autarky or “Sovereignty Theater” — the illusion that installing foreign-designed silicon in domestic server racks creates technological autonomy.

Instead, sovereignty must be evaluated across five functional pillars:

Control, Continuity, Substitutability, Data Authority, and Economic Viability.

In the second installment, Canada’s Sovereign AI Test: Turning Research, Power and Compute into Choice, I applied that framework to Canada’s national capabilities.

Now, we move one level deeper: from the global supply chain and the Canadian state to the company that may become one of the most important instruments of Canada’s sovereign AI ambitions.

Key takeaways

01

A historic investment

Ottawa’s commitment of up to C$1.6 billion forms part of Cohere’s Series E, which targets C$7 billion at a stated valuation of C$35.7 billion. The round has completed only an initial close, and the government’s precise equity stake and governance rights remain undisclosed.

02

A differentiated business model

Cohere’s strength lies in private enterprise deployment. In May 2025, the company told Reuters that approximately 85% of its business came from private deployments.

03

A wider Canadian commitment

Cohere separately expects to invest more than C$4 billion in Canadian research, development, and AI infrastructure through 2031, while creating more than 1,500 Canadian roles.

04

The sovereignty paradox

Open weights, Canadian data centres, and a national AI champion expand Canada’s choices, but none independently guarantees sovereignty. The most persistent dependencies remain in semiconductor supply chains, infrastructure jurisdiction, software, and enterprise retrieval systems.

1. The National Champion: Why Cohere Matters

The global foundation-model market is highly concentrated.

According to Menlo Ventures’ 2025 State of Generative AI in the Enterprise report, Anthropic accounted for an estimated 40% of enterprise LLM API spending, OpenAI 27%, and Google 21%.

Together, these three providers represented approximately 88% of the measured enterprise LLM API market.

These figures are estimates based on a survey of 495 U.S. enterprise AI decision-makers. They are not a measurement of every AI model, deployment, or application worldwide. Nevertheless, they illustrate the concentration of commercial model demand among a small group of predominantly American providers.

For middle powers, this concentration creates a strategic vulnerability.

Businesses and governments may increasingly depend on foreign-controlled models for essential services, institutional knowledge, research, and economic productivity.

Canada faces this challenge while its domestic adoption of AI is accelerating. According to Statistics Canada’s second-quarter 2026 business survey, 19.2% of Canadian businesses reported using AI to produce goods or deliver services during the preceding 12 months, compared with 6.1% in the second quarter of 2024.

The same survey identified cybersecurity and privacy concerns as the most frequently reported barriers to AI adoption, cited by 13.4% of businesses, ahead of cost at 10.6%.

This is important because it identifies a problem Cohere’s private-deployment strategy is specifically designed to address: organizations want the productivity benefits of AI without losing control over their information and systems.

Against this backdrop, Cohere occupies an unusual strategic position.

Founded in Toronto in 2019 by Aidan Gomez, Ivan Zhang, and Nick Frosst, Cohere emerged from Canada’s deep learning research ecosystem.

Gomez was among the co-authors of the seminal 2017 research paper Attention Is All You Need, which introduced the Transformer architecture that underpins modern large language models.

Unlike companies primarily competing for consumer chatbot dominance, Cohere has concentrated on enterprise AI, retrieval-augmented generation (RAG), information retrieval, multilingual capabilities, and deployment arrangements designed for organizations with demanding security and data-governance requirements.

This is a strategically meaningful distinction.

Canada does not need to reproduce every business model developed in Silicon Valley. It needs competitive capabilities in areas where its economic structure, public institutions, research strengths, and relationships with allied countries provide a realistic opportunity.

Cohere’s focus on regulated enterprises and governments gives it a plausible route to international relevance without attempting to outspend the largest American laboratories across every dimension of consumer AI.

A transatlantic sovereign AI alliance

Perhaps the most consequential development in Cohere’s international strategy is its planned combination with Germany’s Aleph Alpha.

On September 16, 2026, the two companies announced that they had signed a definitive business combination agreement, following the initial announcement of their planned combination in April.

The agreement envisions a unified company operating globally as Cohere, with dual headquarters in Toronto and Berlin, while Aleph Alpha’s Heidelberg office will continue as a research centre.

The combined organization is expected to bring together more than 1,000 employees across both continents. Aleph Alpha’s co-CEO Ilhan Scheer and co-founder Samuel Weinbach are also expected to assume senior leadership roles in the combined company upon closing.

The transaction remains subject to final regulatory approvals. It is important to distinguish the signing of a definitive agreement from the completion of the transaction.

The strategic logic extends beyond corporate expansion.

Cohere contributes its enterprise models, retrieval technology, deployment expertise, and North American market presence. Aleph Alpha contributes its European public-sector relationships, industrial partnerships, and experience serving organizations with demanding requirements around data control and technological autonomy.

Germany has sought to establish stronger domestic AI capabilities rather than relying exclusively on foreign foundation-model providers.

The combination creates an opportunity to connect Canadian and European sovereign AI initiatives, potentially offering governments and regulated industries an alternative to dependence on a small number of American technology platforms.

It also introduces a new dimension to middle-power cooperation.

Canada cannot independently reproduce the entire frontier AI ecosystem of the United States or China. Germany faces similar economic constraints. Combining research capabilities, commercial markets, institutional relationships, and computing infrastructure may produce a stronger competitive position than either country could achieve alone.

However, a transatlantic company is not automatically a sovereign one.

Combining Canadian and German corporate interests may diversify institutional relationships, but it does not eliminate dependence on globally concentrated semiconductor manufacturing, American computing platforms, or international capital markets.

The partnership is therefore best understood as an attempt to create greater strategic choice through international cooperation, not as an achievement of technological independence.

Source: Cohere and Aleph Alpha — Definitive business combination agreement, September 16, 2026.

Canada’s C$1.6 billion bet and the Series E context

The October 29 announcement significantly changes the relationship between Cohere and the Canadian state.

The federal government committed to making an equity investment of up to C$1.6 billion in Cohere as part of a much larger Series E financing round.

According to Cohere’s announcement, the round targets up to C$7 billion at a stated valuation of approximately C$35.7 billion. Germany’s Schwarz Group is leading the round, with Bell Ventures and Bullhound Capital among the new investors and the Government of Canada joining alongside returning investors including NVIDIA, Inovia Capital, and Prosperity7 Ventures.

Cohere describes the proposed financing as the largest private capital raise in Canadian history.

However, the round has completed only an initial close. The company has not disclosed the total amount committed, the final closing date, or the amount of the government’s investment that has actually been transferred.

Reuters also reported that the precise size and terms of Ottawa’s equity stake remain undisclosed.

This creates an important distinction between the government’s public commitment and the financial structure of its actual ownership.

The question is not simply how much public capital Canada is prepared to invest.

It is what rights and responsibilities accompany that investment.

The government has an interest in a company’s commercial success as a shareholder. As a policymaker, it also has a responsibility to protect competition, maintain institutional independence, and ensure that national strategic capabilities remain available beyond the lifetime of a single corporation.

These interests may align, but they are not identical.

A government that becomes a shareholder in a major AI company must also consider how its investment interacts with procurement decisions, public-sector adoption, competition policy, and national security requirements.

The investment is therefore a test not only of Cohere’s commercial prospects, but of Canada’s ability to govern the relationship between public capital and private technological power.

Sources: Government of Canada — October 29, 2026 and Cohere — Series E announcement.

A separate C$4 billion commitment to Canada

On October 30, 2026, Cohere announced an expansion of its longer-term Canadian investment programme.

The company expects to invest more than C$4 billion in Canadian research, development, and AI infrastructure between 2024 and 2031.

It also expects to create more than 1,500 Canadian roles, including approximately 1,000 highly skilled technical positions across AI research, engineering, product development, infrastructure, and deployment.

The company identified several initiatives already contributing to this commitment, including:

  • A C$309 million GPU-cloud partnership with Bell AI Fabric.
  • University partnerships, including the University of Toronto and the University of Waterloo.
  • Research collaborations involving Mila, Vector, and Amii.
  • More than 250 paid student opportunities by 2029.

These are company-announced investment and employment projections, not independent guarantees of future delivery.

The C$4 billion programme must also be distinguished from the government’s separate commitments, including its up-to-C$240 million contribution toward the Cambridge compute project and the new C$1.6 billion equity investment.

The initiatives have different scopes and periods. They should not simply be added together as though they represent independent new investment.

Nevertheless, the wider commitment illustrates something important: Cohere’s development is increasingly intertwined with Canada’s talent, research, infrastructure, and commercialization strategies.

Source: Cohere — Advancing its C$4 billion Canadian commitment, October 30, 2026.

Cohere and Canada’s AI for All strategy

The investment should also be understood within the broader context of Canada’s national AI policy.

On June 4, 2026, Prime Minister Mark Carney launched AI for All, Canada’s new national artificial intelligence strategy.

The strategy is organized around six pillars, including:

  • Building the foundations of sovereign Canadian AI.
  • Scaling Canadian AI champions.
  • Developing trusted international partnerships.

Its fifth pillar explicitly identifies government procurement as a strategic anchor for growing Canadian AI companies.

This makes Cohere’s financing particularly significant.

The government is not merely funding a technology company. It is testing an industrial-policy model in which public capital, procurement, domestic infrastructure, and international partnerships are intended to help Canadian AI businesses scale.

Cohere is arguably the first major test of that fifth pillar.

But the strategy also reveals the tension at the centre of this article.

A government acting as investor, strategic customer, and national technology policymaker must ensure that supporting a domestic champion does not unintentionally weaken competition or create another single-vendor dependency.

Source: Government of Canada — AI for All, June 4, 2026.

2. Enterprise Deployment: Control and Data Authority

To understand why Cohere has a differentiated position in the global AI market, it is necessary to look beyond model benchmarks and examine its deployment architecture.

Cohere has built much of its enterprise strategy around private deployment flexibility.

Rather than depending exclusively on a public, multi-tenant API model, it provides organizations with options to operate AI systems within customer-controlled environments.

Its private deployment documentation describes deployment arrangements that include customer-controlled on-premises infrastructure and private cloud environments.

These options matter for financial institutions, hospitals, government departments, defence organizations, and enterprises processing sensitive information.

For such organizations, the question is not simply whether a model produces accurate answers.

It is where data is processed, who can access it, who controls the infrastructure, and whether critical services can continue operating if an external provider becomes unavailable.

The commercial evidence behind private deployment

There is evidence that this strategy has commercial traction.

In May 2025, Reuters reported that approximately 85% of Cohere’s business came from private deployments, following the company’s strategic shift toward customized enterprise solutions in regulated industries.

This was a company-reported figure at that time, not an audited breakdown of Cohere’s current 2026 revenue.

Nevertheless, it provides an important indication of the commercial importance of private deployment to Cohere’s business model.

A bank may be willing to pay for a model that can operate within an environment designed around its security and regulatory requirements. A government department may value local control, auditability, and the ability to restrict external connectivity more than marginal improvements in general-purpose benchmark performance.

Cohere’s competitive differentiation is therefore not simply the intelligence of its models. It is the ability to deliver those models within environments that organizations can govern.

Reuters — Cohere’s enterprise strategy and private deployments, May 15, 2025.

The difference between private deployment and sovereignty

Deployment flexibility should not be confused with complete sovereignty.

A model running inside a Canadian virtual private cloud may still depend on a foreign-controlled cloud operator. An on-premises deployment may provide greater physical and operational control, but it remains dependent on hardware manufacturers, firmware providers, model updates, security support, and specialized technical expertise.

The location of inference is not sufficient evidence of who ultimately controls the system.

A useful distinction emerges between three deployment arrangements:

Public API: The organization primarily depends on an external model provider and its service availability, subject to contractual and technical controls.

Private cloud or VPC: The organization gains greater control over the deployment environment and data isolation, but may remain dependent on the cloud operator’s legal jurisdiction and operational authority.

On-premises or air-gapped deployment: The organization controls more of the physical and operational environment, but must independently manage hardware, software, maintenance, security, and continuity.

Each arrangement offers a different balance of control, cost, and operational complexity.

The CLOUD Act and juridical sovereignty

The legal dimension is particularly important.

The United States CLOUD Act allows American authorities, subject to applicable legal process, to seek certain data within the possession, custody, or control of providers subject to U.S. jurisdiction, regardless of where that data is physically stored.

This does not mean American authorities have unrestricted direct access to every foreign-hosted server.

The Government of Canada’s own Digital Sovereignty Framework recognizes that using a Canadian supplier or storing information in Canada does not automatically remove foreign-jurisdiction exposure.

The same framework also provides an important qualification: its authors found no documented cases of foreign governments seeking access to Canadian enterprises’ data held by suppliers.

This distinction matters.

The exposure is structural legal reach, not evidence that foreign governments routinely access Canadian enterprise information through commercial cloud providers.

It is precisely why a risk-based approach is preferable to treating every foreign provider as inherently untrustworthy.

A related distinction is developed in the Munk School’s March 2026 report, Sovereign by Design: Strategic Options for Canadian AI Sovereignty.

The report distinguishes between:

  • Juridical cloud sovereignty: Canadian-owned operators with a legal separation from foreign jurisdictions.
  • Contractual cloud sovereignty: Achieving defined sovereignty outcomes through contractual and technical controls that any provider can meet.

The appropriate arrangement depends on the sensitivity of the workload.

For highly sensitive workloads, the strongest arrangement may be a Canadian-operated environment in which infrastructure ownership, operational authority, data governance, and legal jurisdiction are aligned.

For less sensitive workloads, a carefully negotiated foreign-operated environment may be economically reasonable.

Sovereignty should therefore be determined by the sensitivity and strategic importance of the workload, not by a simplistic assumption that every foreign dependency must be eliminated.

Intelligence applications: the CSE relationship

Cohere’s engagement with Canada’s national security establishment provides another important dimension to its public-sector strategy.

In June 2025, The Logic reported that Cohere had a contract with the Communications Security Establishment (CSE), Canada’s national cryptologic and cybersecurity agency.

The reporting was based on government documents obtained through an access-to-information request. An October 1, 2024 briefing memorandum identified a CSE contract for Cohere technology.

The memorandum also described Cohere’s models as deployable at client premises and air-gapped from the original model. This is particularly relevant to the deployment architecture discussed in this section.

The nature, scope, and financial value of the contract were not disclosed. CSE declined to comment on its contracts, citing national security.

This is meaningful evidence of a government relationship, but it should not be overstated. The public information does not establish which specific Cohere technology was supplied, how it was deployed, or whether it was used in an operational intelligence environment.

Nevertheless, the existence of the contract demonstrates that Cohere’s public-sector engagement extends beyond general government innovation initiatives into Canada’s national security ecosystem.

The Logic — Cohere had a contract with Canadian cyber intelligence agency, June 11, 2025.

Defence applications: moving beyond theory

Cohere’s relationship with Thales Canada provides a more visible example of the transition from enterprise AI development to defence applications.

In December 2025, the two companies signed a partnership agreement for advanced AI solutions in naval and maritime in-service support.

A more substantial milestone followed on October 6, 2026.

Thales Canada and Cohere announced a new task order under the Arctic and Offshore Patrol Vessels and Joint Support Ships In-Service Support (AJISS) contract.

The initiative represents the first integration of agentic AI into the Royal Canadian Navy’s maintenance and support activities through this collaboration.

The announced solution combines Thales’ defence and digital technology expertise, its Montréal-based cortAIx AI accelerator for critical systems, and Cohere’s enterprise AI capabilities.

The objectives include reducing administrative burdens, accelerating maintenance and logistics processes, and improving ship readiness and availability.

Importantly, the announced approach includes human oversight.

Cohere co-founder Ivan Zhang also stated that the technology is designed to operate in disconnected environments while keeping Royal Canadian Navy personnel firmly in control.

That last detail is particularly relevant to sovereignty.

A naval vessel cannot assume continuous access to a commercial cloud service. A mission-critical system must be able to operate in environments where connectivity is limited, unavailable, or deliberately restricted.

This is a meaningful architectural capability. However, the announcement does not establish that the system is already operational across the fleet, nor does it disclose whether the solution will operate aboard vessels, at shore facilities, or through a hybrid deployment.

Nevertheless, this development demonstrates a concrete pathway through which a Canadian AI developer can contribute to a strategic defence capability alongside an established defence contractor.

It also raises an important governance question: when AI systems begin coordinating workflows in national defence environments, who determines their permitted actions, who audits their decisions, and who remains accountable?

Thales and Cohere — Royal Canadian Navy AI milestone, October 6, 2026.

Building domestic deployment channels

Another important development is the collaboration between Bell AI Fabric, Cohere, Hypertec, and BUZZ High Performance Computing.

The partnership brings together complementary Canadian capabilities:

  • Bell provides data-centre infrastructure and connectivity.
  • Cohere contributes enterprise AI models and software capabilities.
  • Hypertec supplies Canadian-manufactured hardware.
  • BUZZ HPC provides scalable accelerated-computing infrastructure.

The companies intend to support advanced AI research and deployment using Canadian infrastructure and commercial partners.

This is more meaningful than simply describing a domestic data centre.

It demonstrates how Canadian organizations can participate in different layers of an AI infrastructure stack rather than relying entirely on foreign hyperscalers.

Nevertheless, the presence of Canadian partners does not eliminate dependence on foreign technologies, including NVIDIA’s computing platform.

The strategic opportunity is not complete technological self-sufficiency. It is the development of sufficient domestic capability to maintain choice, control, and resilience when international conditions change.

Source: Bell AI Fabric and Cohere partnership.

3. Compute Infrastructure: The Reality of Managed Interdependence

If Cohere’s deployment strategy demonstrates how to strengthen Control and Data Authority at the inference layer, its compute expansion illustrates the limits of domestic sovereignty in AI training.

On December 6, 2024, the Government of Canada announced an investment of up to C$240 million in Cohere through its Canadian Sovereign AI Compute Strategy.

The investment was finalized on March 20, 2025, as part of Cohere’s C$725 million compute expansion project in Cambridge, Ontario.

The project is being developed in partnership with CoreWeave, the American AI infrastructure provider that will build and operate the facility. Cohere will act as its anchor tenant.

This project is an important component of Canada’s sovereign AI ambitions.

It also illustrates why infrastructure location and infrastructure control must be evaluated separately.

A data centre may be physically located in Canada and support Canadian AI development while still depending on foreign-controlled operational capabilities and internationally concentrated hardware supply chains.

Understanding who builds, operates, and finances the infrastructure

It is important to distinguish the different roles in this project.

The federal government’s project documentation identifies CoreWeave as the infrastructure partner responsible for building and operating the Cohere compute facility.

A separate development in Cambridge has also attracted attention because CPP Investments announced C$225 million in construction financing for a 54 MW hyperscale data-centre expansion in July 2025.

That separate financing arrangement involves a joint venture between Related Digital, TowerBrook Capital Partners, and Ascent, alongside Deutsche Bank financing. The announcement describes an unnamed GPU-focused AI cloud provider as the long-term tenant.

The public financing announcement does not identify Cohere as the borrower or beneficiary. It should therefore not be presented as C$225 million of financing for Cohere’s C$725 million project.

These are distinct infrastructure transactions, even though both relate to Cambridge’s growing data-centre market.

This distinction is important because sovereign AI discussions frequently conflate ownership, construction, operation, financing, and tenancy.

They are different forms of control.

A company may own a model, another may own or finance the building, a third may operate the cloud platform, and a fourth may manufacture the accelerators.

The sovereignty analysis must account for all of them.

1. Training scale versus sensitive inference

Training a competitive foundation model requires large quantities of specialized accelerators, high-speed networking, advanced cooling, distributed storage, and sophisticated software orchestration.

These capabilities are expensive and difficult to establish.

For Canada, building every component domestically before participating in frontier AI development would create an enormous economic and strategic disadvantage.

Partnering with CoreWeave gives Cohere access to advanced computing infrastructure in Canada without requiring the country to independently develop a globally competitive GPU-cloud operator first.

However, the requirements for model training are not identical to those for sensitive inference.

Training workloads may involve public, licensed, or internally curated datasets. Inference workloads may process classified government information, confidential commercial records, health information, and other sensitive institutional data.

The appropriate sovereignty requirements therefore differ.

Canada may reasonably accept foreign infrastructure operators for some training workloads while insisting on substantially stronger domestic operational and jurisdictional control for its most sensitive inference workloads.

2. Contractual versus juridical sovereignty

The Cambridge project represents a form of managed interdependence.

Cohere gains access to computing resources located in Canada, while the facility’s American operator remains subject to the legal and regulatory environment of its home jurisdiction.

Contractual safeguards, encryption, access controls, customer-held keys, and operational procedures can reduce exposure.

But contracts cannot entirely erase the legal obligations of the infrastructure provider.

This is why contractual and juridical sovereignty should be treated as distinct concepts.

For sensitive workloads, Canada needs infrastructure operators capable of providing meaningful domestic legal and operational control.

Companies such as Hypertec, ThinkOn, Aptum, and other Canadian infrastructure providers can contribute to that ecosystem, alongside government-operated infrastructure.

The objective should not be to prohibit international partnerships. It should be to ensure that Canada has credible alternatives when foreign dependencies become unacceptable.

3. Semiconductor and software chokepoints

Even with servers located in Cambridge, the advanced computing ecosystem remains dependent on a global semiconductor supply chain.

NVIDIA’s accelerator architecture, proprietary CUDA software ecosystem, networking technologies, advanced packaging, and manufacturing relationships are central to contemporary AI infrastructure.

The leading-edge semiconductor fabrication industry is also geographically concentrated, with Taiwan’s TSMC occupying a particularly important position.

These dependencies cannot be eliminated simply by establishing Canadian data centres.

Export controls, regulatory restrictions, geopolitical disruptions, or changes in commercial licensing could affect access to critical computing technologies.

The January 2025 U.S. AI Diffusion Rule, followed by the U.S. government’s May 2025 announcement that it would stop enforcing the rule and move to rescind it, demonstrated how quickly the regulatory environment surrounding advanced AI computing can change.

The lesson is not that Canada must manufacture every semiconductor domestically.

It is that Canada must understand its exposure to external chokepoints and establish alternatives wherever those dependencies threaten critical national capabilities.

This is managed interdependence: accepting participation in a global technology ecosystem while ensuring that no single external provider can dictate the country’s strategic options.

The missing ingredient: Continuity

A Canadian data centre is valuable, but infrastructure location alone cannot guarantee continuity.

The real question is whether Canadian organizations can maintain access to critical compute resources when commercial relationships, export rules, or geopolitical conditions change.

This requires several layers of resilience:

  • Multiple infrastructure providers.
  • Alternative accelerator and software configurations where technically feasible.
  • Long-term access and maintenance arrangements.
  • Tested workload migration procedures.
  • Independent operational expertise.
  • Continuity plans for critical model-serving environments.

The country does not need to duplicate the entire global hardware supply chain. It does need to avoid creating critical dependencies without viable contingencies.

Government of Canada — Original C$240 million commitment, December 6, 2024.

Government of Canada — Finalized Cohere investment, March 20, 2025.

4. Open Weights and the Hidden Problem of Substitutability

One of the most strategically important developments in Cohere’s model portfolio is the release of Command A+ on May 20, 2026.

According to Cohere’s technical documentation, Command A+ is a sparse Mixture-of-Experts (MoE) model containing:

  • 218 billion total parameters.
  • Approximately 25 billion active parameters per token.
  • A 128,000-token context window.
  • Up to 64,000 output tokens.
  • Text and image input capabilities.
  • Multilingual support across 48 languages.
  • Reasoning and agentic tool-use capabilities.
  • An Apache 2.0 licence for its released W4A4 quantized checkpoint.

The model’s architecture illustrates an important principle of modern AI economics: total parameter count does not necessarily determine the computing resources required for every inference operation.

In a sparse MoE model, only selected expert components are activated for a given token. This can make a model with hundreds of billions of parameters more practical to deploy than a similarly sized dense architecture.

Cohere documents W4A4 four-bit weight-and-activation quantization configurations using a single NVIDIA B200 GPU or two H100 GPUs.

This is strategically relevant.

A model that requires an enormous multi-node cluster limits deployment to a small number of wealthy organizations and hyperscale infrastructure providers.

Reducing hardware requirements makes advanced capabilities more accessible to universities, research organizations, enterprises, and public institutions with their own computing infrastructure.

It does not make frontier-grade AI inexpensive, nor does it eliminate hardware dependency. H100 and B200 accelerators remain high-end data-centre components.

But it lowers the infrastructure threshold compared with systems requiring much larger clusters.

Open weights are not the same as fully open-source AI

There is another important distinction.

Command A+ is distributed under the Apache 2.0 licence for the documented W4A4 quantized checkpoint.

This provides users with significant freedom to obtain, deploy, modify, and redistribute that released artifact, subject to the licence’s conditions.

However, open weights do not automatically constitute a fully open-source AI system.

The Open Source Initiative’s Open Source AI Definition identifies three important components required for meaningful modification of an AI system:

  • Information about the training data.
  • The complete code used to process data, train, and run the system.
  • Model parameters, including weights.

Cohere also publishes an EU AI Act public summary of Command A+’s training content, which provides transparency about the data sources and processes used to develop the model.

However, a regulatory training-content summary and the complete information needed to independently reproduce an AI system are not necessarily the same thing.

This distinction matters for governments and enterprises evaluating long-term technological control.

The ability to download a model is not identical to the ability to recreate its training process, maintain its software ecosystem, or independently reproduce its capabilities.

The RAG portability problem: replacing the model is not enough

There is a deeper technical challenge that is frequently overlooked in discussions of AI sovereignty.

Most enterprise AI systems are not simply foundation models. They are interconnected systems built around institutional data, embedding models, vector databases, retrieval engines, reranking systems, orchestration frameworks, and applications.

Consider a typical retrieval-augmented generation architecture.

An organization takes its documents, policies, contracts, research records, and other institutional knowledge and converts them into numerical representations called embeddings.

These embeddings are stored in a vector database, which allows the system to retrieve relevant information when a user submits a question.

A foundation model then uses the retrieved information to generate a response.

The challenge emerges when an organization tries to change one of these components.

Replacing the foundation model does not necessarily require rebuilding the knowledge index. Replacing the embedding model often does.

Embeddings generated by different models generally occupy different vector spaces. Even when two models produce vectors of identical dimensions, those vectors are not automatically interchangeable.

MongoDB’s documentation for its Voyage AI embedding models explains why changing embedding models requires regenerating corpus embeddings to preserve reliable retrieval.

For a large enterprise, this can introduce significant migration costs, operational disruption, and validation requirements.

This means an organization can successfully move away from a foreign foundation-model provider while remaining locked into proprietary retrieval infrastructure.

The problem extends beyond embeddings.

A complete migration may also involve:

  • Vector database formats and export capabilities.
  • Proprietary reranking models.
  • Retrieval and orchestration frameworks.
  • Application-specific model integrations.
  • Metadata and document-permission structures.
  • Evaluation pipelines and quality benchmarks.

The implication is important:

True substitutability must extend across the entire enterprise AI stack, not merely the foundation-model layer.

Cohere’s own retrieval stack: Embed 5 and Rerank 5

This challenge applies to Cohere itself.

In September 2026, Cohere released Embed 5, a new family of embedding models designed for enterprise retrieval across text, images, and complex documents.

The family includes Pro and Fast variants that share an embedding space, allowing an organization to index a corpus using one variant and query it using the other.

Cohere provides Embed 5 through its API, partner platforms, and single-tenant Model Vault deployments.

On October 21, Cohere introduced Rerank 5, its latest enterprise reranking models.

Rerank 5 is designed to improve retrieval quality by reordering candidate documents according to their relevance to a query. Its Context Focus capability can identify relevant passages within documents, helping reduce the amount of unnecessary information passed to the foundation model.

Cohere documents deployment through its API and Model Vault, with private VPC and on-premises deployment available through sales arrangements.

These releases strengthen Cohere’s position beyond the foundation-model layer.

But they also introduce a procurement question that applies to every provider, including Canada’s national champion:

Can customers independently maintain, export, migrate, and replace the retrieval components on which their institutional knowledge depends?

Model Vault, for example, is a Cohere-managed, isolated single-tenant inference environment. It offers stronger isolation than a shared API but should not automatically be confused with customer-operated infrastructure.

Buyers should establish in writing which components they can export, which they can self-host, and what migration support and exit rights they retain.

Sources: Cohere Embed 5 documentation and Cohere Rerank 5 documentation.

Standardization as an exit strategy

There are practical ways to reduce this dependency.

Open inference frameworks such as vLLM provide OpenAI-compatible APIs for serving supported models.

These interfaces can reduce the integration effort involved in moving between compatible model providers.

However, API compatibility alone does not guarantee equivalent model behaviour, performance, tool use, or output quality.

Similarly, standardized data-export formats, portable vector indexes where supported, and documented embedding migration procedures can reduce the costs of changing retrieval providers.

Governments and enterprises should require explicit portability provisions and realistic migration plans in their AI procurement agreements.

For embedding systems, this should include documented model versions, exportable source documents and metadata, tested re-embedding procedures, and independent evaluation of retrieval quality after migration.

A national AI strategy that makes foundation models replaceable but leaves enterprise knowledge infrastructure locked to a single provider has only solved part of the problem.

For Canada, this is where Cohere’s open-weight strategy becomes particularly interesting.

Its long-term sovereignty value will depend not just on whether Command A+ can run outside Cohere’s own infrastructure, but also on whether institutions can integrate it into systems they fully understand, maintain, and replace.

5. Safe-by-Design AI: Connecting Commercial Capability With Scientific Responsibility

Technological sovereignty is not simply about who owns the models or controls the computing infrastructure.

It is also about who establishes the standards governing how these systems behave.

As AI models become more capable of planning, interacting with external tools, and executing multi-step tasks, the distinction between generating information and taking consequential action becomes increasingly important.

An AI system that summarizes a government document creates one category of risk.

An autonomous system authorized to initiate financial transactions, modify critical infrastructure configurations, or execute administrative decisions introduces a different category entirely.

Canada’s AI ecosystem has an opportunity to contribute to this challenge through independent safety research.

One important initiative is Montréal-based LawZero, a non-profit AI safety research organization founded by Yoshua Bengio.

On September 16, 2026, Canada and Germany announced plans to invest C$150 million and €100 million, respectively, in LawZero — roughly C$300 million combined.

These are announced funding commitments, not proof that the entire amount has already been disbursed.

The initiative aims to advance a different approach to the development of capable AI systems, emphasizing safety, transparency, and scientific foundations.

A relevant scientific contribution is the June 2026 paper, Safety from Honesty in a Disinterested AI Predictor, by Bengio and collaborators.

The research proposes a Scientist AI Predictor designed to estimate facts, actions, and consequences without itself adopting the goal-directed behaviour associated with an autonomous agent.

Its safety argument relies on a particular training objective and assumptions about the system’s behaviour and the probability of dangerous predictors emerging.

This is an important research direction, but it should not be interpreted as proof that deployed AI systems can already be guaranteed safe.

Formal results depend on their assumptions, and the use of a safe predictor inside a broader agentic system can introduce additional risks through the surrounding architecture.

The strategic opportunity for Canada is to connect independent research efforts with its commercial AI ecosystem.

Cohere develops enterprise AI products that can be deployed in sensitive environments. LawZero explores alternative approaches to safety and the development of more trustworthy AI. Canada’s National AI Institutes — Mila, the Vector Institute, and Amii — contribute research expertise and talent development.

These organizations should not be treated as a single integrated system. Their independence is valuable.

A healthy national AI ecosystem needs commercial innovation, academic research, independent evaluation, and public accountability.

The objective should be to make trustworthy AI an independently verifiable capability, rather than a marketing claim attached to a domestic model.

Government of Canada — Canada and Germany’s LawZero funding announcement, September 16, 2026.

6. The Limits of the Sovereign Champion

Cohere demonstrates that a middle power does not need to reproduce every element of the global AI supply chain to develop strategically important capabilities.

However, the emergence of a successful national champion also creates new questions.

There are five dependencies that Canada must continue to address.

Hardware dependency. Cohere can own model weights without owning the global semiconductor ecosystem required to train and operate them. Advanced accelerators, firmware, networking, and specialized software remain concentrated in international supply chains.

Infrastructure dependency. A Canadian data centre operated by a foreign provider may provide domestic compute capacity without providing complete domestic operational or jurisdictional control.

Commercial dependency. A company’s ability to compete internationally depends on sustainable revenue, margins, capital availability, and continuous investment in research. Government investment can accelerate growth, but it cannot substitute indefinitely for commercial viability.

Ecosystem dependency. Open weights make it easier to change a model provider, but they do not automatically eliminate dependence on specialized inference software, cloud orchestration, hardware compatibility, or technical support.

Governance dependency. When government funding, procurement, research priorities, and critical infrastructure become concentrated around one private corporation, national resilience may gradually become dependent on corporate decision-making.

These are not arguments against Cohere.

They are reasons to avoid confusing the success of one organization with the sovereignty of an entire country.

The National Champion Paradox

The October 29 investment makes this distinction particularly important.

Reuters reported that Cohere told investors its annual recurring revenue had grown 165% year over year as of the third quarter of 2026, while its customer count had grown approximately twentyfold to around 300.

These are company-reported commercial indicators, not independently audited financial statements. They demonstrate momentum, but they do not establish profitability or long-term financial sustainability.

The economics of advanced AI remain demanding. Companies must continually invest in computing infrastructure, talent, research, product development, and international expansion.

A government equity investment can help a domestic company compete at a scale otherwise difficult to achieve.

But the government should measure success not only by Cohere’s valuation, revenue, or international market share.

It should also evaluate whether the investment increases the number of organizations capable of building, deploying, operating, securing, and independently evaluating AI systems in Canada.

A national champion should strengthen the ecosystem around it, not replace that ecosystem.

The concentration of relationships: a case for independent evaluation

There is another dimension worth considering.

Look at the organizations surrounding Cohere’s development.

NVIDIA is both an investor and a supplier of the advanced accelerators on which Command A+ is documented to run.

Bell is an investor, a strategic infrastructure partner, and a provider of Canadian AI computing capacity.

Schwarz Group leads the Series E and has also been a significant supporter of Aleph Alpha.

These relationships are not inherently problematic. Small and emerging technology ecosystems naturally develop close commercial relationships because the number of organizations capable of participating at the frontier is limited.

However, the concentration of financial, infrastructure, and commercial relationships makes independent evaluation increasingly important.

The same principle applies to the policy research informing Canada’s sovereign AI debate.

The Munk School’s March 2026 report, Sovereign by Design: Strategic Options for Canadian AI Sovereignty, acknowledges financial support from RBC, Cohere, Bell Canada, OpenText, and Ontario Teachers’ Pension Plan.

The authors explicitly state that the Munk School maintained full editorial independence and that the funders did not influence the report’s findings.

That disclosure does not invalidate the research. It is important context for readers evaluating policy recommendations involving organizations that also participate in the Canadian AI ecosystem.

The report itself recommends ongoing validation of providers’ sovereignty claims rather than relying on one-time certification.

Canada needs precisely this kind of independent assessment capability.

Sovereign AI cannot depend entirely on companies evaluating their own security claims, governments evaluating their own investments, or policy research without transparent funding disclosures.

The country must develop independent technical expertise capable of examining architecture, infrastructure ownership, legal jurisdiction, operational continuity, and migration capabilities.

The policy paradox: Canadian champion versus single-vendor dependency

The Munk School report also identifies an important tension in the debate over national AI champions.

Among its policy options, the report proposes establishing procurement preferences for Canadian model providers where possible and mandating Canadian models for certain higher-sensitivity workloads.

It also identifies Cohere as the only Canadian foundation-model option in its assessment of the market, while recognizing the concentration risk created by a single domestic provider.

This creates a difficult policy question.

A preference for Canadian models may accelerate domestic commercialization and ensure that Canadian institutions have opportunities to develop local capabilities.

But mandating a Canadian foundation model in a market with only one major domestic provider risks creating a single-vendor procurement structure.

A policy designed to reduce foreign dependency could therefore introduce a different form of dependency at home.

The way forward is not to abandon support for Canadian models. It is to make that support conditional on competition, portability, transparent technical standards, and credible alternatives.

Open-weight models, standardized interfaces, containerized deployment, and demonstrated migration capabilities can provide a strategic hedge.

The objective should be to support Cohere without making Cohere the only possible answer.

That is the distinction between building a national champion and building a resilient national ecosystem.
7. From National Champion to Ecosystem Anchor: What Canada Must Do

For Cohere’s capabilities to translate into durable national sovereignty, Canadian policymakers and enterprise leaders need to address several structural challenges.

1. Make public investment conditional on measurable national capability

The government’s staged equity investment creates an opportunity to establish clear accountability.

The commitments to Canadian jobs, infrastructure, research partnerships, and domestic corporate presence should be measured transparently.

But the evaluation should extend beyond corporate growth.

Canada should also assess whether the investment creates additional domestic engineering expertise, research capability, independent infrastructure options, and commercially sustainable suppliers.

Public investment should produce long-term national assets, not simply a more valuable private corporation.

2. Pool public and enterprise demand without creating vendor lock-in

The Canadian federal and provincial governments represent a substantial potential market for domestic AI solutions.

Healthcare, education, public administration, research institutions, and financial services all have use cases where secure and locally governed AI infrastructure may provide strategic value.

However, governments should not solve their dependency on foreign AI vendors by creating exclusive dependence on one domestic vendor.

Procurement should encourage competition among Cohere, Canadian cloud providers, alternative model developers, research organizations, and qualified international partners.

There is also a commercial opportunity beyond government.

Canadian application-layer companies such as Shopify, Clio, Wealthsimple, OpenText, Ada, and BenchSci have developed products and services with international reach.

These businesses represent potential sources of sustained demand for secure enterprise AI capabilities.

Encouraging domestic application companies to evaluate Canadian models and infrastructure for suitable production workloads could create a stronger commercial relationship between Canadian application developers, cloud operators, and model providers.

The objective should be to strengthen domestic capabilities while preserving the ability to choose among competing technologies.

3. Address Shadow AI: Demand already exists

Canada does not necessarily need to create demand for enterprise AI from scratch.

A substantial portion of the workforce is already using AI, often faster than employers can establish appropriate enterprise solutions.

An IBM-commissioned study published on September 3, 2025, found that 79% of surveyed full-time office workers in Canada reported using AI at work, while only 25% relied on enterprise-grade AI tools.

The study covered approximately 4,000 office workers across Canada, the United States, Mexico, and Brazil. IBM did not disclose the Canadian sample size, so the figures should be interpreted as survey findings rather than a complete national workforce measurement.

The study described this growing gap as Shadow AI: the use of AI tools and applications in the workplace without formal approval or oversight from an organization’s IT department.

Employees may increasingly use consumer AI services while organizational policies, data-governance controls, and approved enterprise solutions lag behind.

This can create risks involving confidential information, privacy, regulatory compliance, and institutional data governance.

The implication for Canadian AI sovereignty is significant.

Domestic AI infrastructure and enterprise models need to compete not only with foreign enterprise platforms, but also with the convenience of consumer AI applications.

Sovereignty will remain an abstract policy objective unless secure Canadian capabilities become accessible, useful, and economically attractive to ordinary businesses and employees.

IBM Canada — Shadow AI study, September 3, 2025.

4. Make portability a procurement requirement

Government and enterprise buyers should insist on architectures that allow models, data, and application workflows to move between providers.

This requires more than contractual promises.

It requires:

  • Standardized model-serving interfaces.
  • Portable retrieval and orchestration components.
  • Documented embedding and vector-index migration procedures.
  • Customer-controlled encryption keys.
  • Exportable institutional data and metadata.
  • Independent model evaluation.
  • Tested exit and continuity plans.
  • Defined migration timelines and service-level commitments.

Portability must be demonstrated, not merely promised.

A practical test would be to require critical institutions to periodically demonstrate that they can move an application workload to another model and infrastructure provider within an acceptable operational window.

For large RAG systems, this should include the ability to export source documents, regenerate embeddings, rebuild retrieval indexes, and validate retrieval performance after migration.

A system that can only be replaced in theory is not genuinely substitutable.

5. Establish sovereignty requirements based on workload sensitivity

Not every AI workload needs the same degree of domestic control.

Canada should distinguish between three practical categories:

Low-sensitivity workloads: International cloud infrastructure and foreign models may be economically reasonable, provided that privacy, security, and contractual requirements are satisfied.

Strategic workloads: These require stronger contractual protections, customer-controlled environments, documented dependencies, and meaningful continuity arrangements.

Critical workloads: These may require Canadian-operated infrastructure, appropriate domestic legal and operational control, independently assessed security, and restricted external dependencies.

These are conceptual categories in this article, not a reproduction of the Munk School’s numbered Data Sensitivity Spectrum.

The Munk report uses three different classifications: Tier 1 for classified information, Tier 2 for personal and sensitive information, and Tier 3 for general operations. It also varies recommended minimum sovereignty levels according to institutional context.

The important principle is to calibrate sovereignty requirements to the harm that could result from losing control of a particular workload.

The goal is not to build an entirely isolated Canadian AI stack.

It is to ensure that the country’s most important institutions have the level of control, continuity, and substitutability their responsibilities require.

6. Build an economic flywheel around Canadian AI

Canada’s opportunity does not end with foundation models.

A sustainable ecosystem needs commercial demand, domestic infrastructure, technical expertise, and continuous research investment.

Consider the relationship:

Canadian application companies create demand.

Canadian cloud and compute providers supply deployment infrastructure.

Canadian model developers provide foundation models and related AI services.

Research institutions contribute talent, scientific innovation, and independent evaluation.

As these relationships mature, each layer creates demand for the others.

For example, a Canadian financial technology company deploying AI in customer support, compliance, or document analysis could evaluate domestic foundation models running on infrastructure operated by Canadian providers.

The application company gains deployment choices. The model developer gains a commercial customer. The infrastructure operator gains utilization. The broader ecosystem gains engineering experience and recurring demand.

This creates an economic feedback loop.

It is also a more realistic route to sustainable AI sovereignty than relying indefinitely on public subsidies to support isolated technology projects.

The federal and provincial governments can help accelerate this process by becoming anchor customers, pooling appropriate workloads, supporting shared infrastructure, and encouraging competition among qualified domestic providers.

The objective is not to force every Canadian company to adopt Canadian AI.

It is to ensure that Canadian solutions are genuinely competitive options.

8. The Verdict: Cohere Is a Strategic Option, Not the Definition of Sovereignty

Cohere’s position is an important test of whether a middle power can build a globally competitive AI company while operating within an international technology ecosystem dominated by much larger economies.

Its enterprise-focused strategy, private deployment architecture, international partnerships, Canadian compute investment, and open-weight models provide Canada with meaningful strategic options.

The planned combination with Aleph Alpha creates an opportunity to connect Canadian and European AI capabilities.

The Bell partnership demonstrates how domestic infrastructure providers can collaborate with model developers and hardware manufacturers.

The Cambridge project illustrates how Canada can secure access to advanced computing resources while managing foreign infrastructure dependencies.

Command A+ demonstrates how more efficient open-weight models can lower the infrastructure threshold for deploying advanced AI.

Embed 5 and Rerank 5 demonstrate how Cohere is expanding into the retrieval layer that underpins enterprise knowledge systems.

The CSE relationship and the Royal Canadian Navy’s new agentic AI integration initiative provide concrete examples of Cohere’s growing engagement with national security and defence-related applications.

The LawZero initiative shows how Canada can contribute to independent scientific approaches to AI safety.

And the October 29 investment represents a major commitment by the Canadian government to help these capabilities scale.

But none of these developments, individually or collectively, proves that Canada has achieved AI sovereignty.

A domestic model does not eliminate hardware dependency.

An open-weight release does not guarantee operational continuity.

A Canadian data centre does not automatically provide complete juridical control.

A government equity investment does not guarantee a resilient national ecosystem.

And a successful national champion does not remove the need for independent alternatives.

Ultimately, Cohere’s contribution to Canada’s sovereignty should be measured by something more important than its ability to compete with American AI laboratories.

It should be measured by its ability to help create an ecosystem in which Canadian institutions can develop, deploy, govern, audit, and replace critical AI capabilities without becoming dependent on any single corporation, infrastructure operator, or foreign policy decision.

Cohere is not Canadian AI sovereignty. It is an instrument through which Canada may develop greater strategic choice.

And that distinction matters.

The ultimate test of a sovereign AI strategy is not whether a country has a national champion.

It is whether that country can remain operational, competitive, and independent even when its national champion is no longer available.

If Cohere helps Canada achieve that outcome, its greatest contribution will not simply be that it builds Canadian AI models.

It will be that it helps make Canada less dependent on any one model provider — including itself.

What to Watch Next

The success of this strategy will depend on several developments that remain unresolved:

  • Cohere–Aleph Alpha: Completion of the business combination, regulatory approvals, and the actual governance arrangements of the transatlantic company.
  • The Series E: The final financing close and disclosure of the size, terms, and holding structure of Canada’s equity participation.
  • Public investment accountability: Progress against Cohere’s Canadian employment, research, and computing commitments.
  • Retrieval portability: Whether Embed 5, Rerank 5, and other enterprise components provide customers with meaningful exit and migration capabilities.
  • Independent sovereignty assessments: Whether Canada develops the institutional capacity to verify providers’ security, control, continuity, and substitutability claims.

These developments will reveal whether Canada’s investment is creating durable national capabilities or simply increasing the scale of one private company.

Series Navigation and Further Reading

This article is the third installment in a three-part series examining the technology, economics, and governance of sovereign AI.

Part 1 — Global Framework

The Sovereign AI Paradox: Owning the Infrastructure Is Not the Same as Controlling the Technology

An examination of why buying local GPUs without governing models, software orchestration, and extraterritorial legal exposure leads to “Sovereignty Theater,” introducing a five-pillar framework for measuring genuine strategic autonomy.

Part 2 — National Case Study

Canada’s Sovereign AI Test: Turning Research, Power and Compute into Choice

A layer-by-layer analysis of how Canada can convert its research legacy, clean electricity advantage, and federal compute investments into practical operational choice.

Additional research and references

Canadian policy and infrastructure

  • Government of Canada — Cohere investment, October 29, 2026
  • Government of Canada — AI for All, June 2026
  • Government of Canada — Digital Sovereignty Framework
  • Munk School — Sovereign by Design: Strategic Options for Canadian AI Sovereignty, March 2026
  • Government of Canada — Cohere compute expansion project
  • CPP Investments — C$225 million construction financing for a separate Cambridge data-centre expansion

Cohere, intelligence and defence applications

  • The Logic — Cohere had a contract with Canadian cyber intelligence agency, June 2025
  • Thales and Cohere — Royal Canadian Navy AI milestone, October 6, 2026
  • Cohere — C$4 billion Canadian investment commitment, October 30, 2026
  • Cohere — Command A+ technical documentation
  • Cohere — Embed 5 release notes
  • Cohere — Rerank 5 release notes
  • Cohere — Model Vault
  • Cohere and Aleph Alpha — Definitive Business Combination Agreement, September 2026
  • Cohere and Bell — Sovereign AI partnership

Scientific and technical foundations

  • Attention Is All You Need — Vaswani et al., 2017
  • Open Source AI Definition — Open Source Initiative
  • ALCE — Enabling Large Language Models to Generate Text with Citations
  • Safety from Honesty in a Disinterested AI Predictor — Bengio et al., 2026
  • MongoDB — Migrating to a New Embedding Model
  • vLLM — OpenAI-Compatible Server Documentation

Industry context

  • Menlo Ventures — 2025 State of Generative AI in the Enterprise
  • Reuters — Cohere’s enterprise deployment strategy, May 2025
  • Reuters — Cohere’s October 2026 funding round and financial performance
  • IBM Canada — Shadow AI study, September 2025
  • Statistics Canada — AI use by Canadian businesses, Q2 2026

Research disclosure: The Munk School’s March 2026 report acknowledges financial support from RBC, Cohere, Bell Canada, OpenText, and Ontario Teachers’ Pension Plan. The authors state that the Munk School maintained full editorial independence and that funders did not influence the report’s content or findings. This disclosure is included to give readers appropriate context when considering its policy recommendations.

From framework to country to company.

Read the foundational framework and the Canada country analysis alongside this Cohere case study.