Cohere's sales of secure, private AI to large organizations and governments keep growing through 2027 and remain its main source of growth.

67.5% stated confidence, Likely (55–80%). Starting point on the record’s first day: Likely (55–80%), 67.5 per cent, the average of five readings, all five of Likely. Attention level: For information. Trend: Steady.

The judgment as stated before the record began

Cohere sells AI that runs where the customer decides: in the customer's own data centre, in a private cloud or fully cut off from the internet. In May 2025 the company told Reuters that most of its business came from such private deployments, and its products since then point the same way. Its customers and partners include banks, telecommunications companies, health and life-sciences firms, defence suppliers and governments, which pay for control over their data as much as for capability. Its chief executive said in December 2025 that spending billions more each year to improve a model would not earn back its cost. The assessment would weaken if its sales of private AI stopped growing as customers moved to larger rivals, or if Cohere's growth came mainly from the businesses it buys, from use through shared public clouds, or from a turn to spending on computing at the scale of the largest AI labs.

How the record started this assessment

On 15 October 2026, before the record began, a model read the 10 developments below five separate times, first reported in the three weeks before and linked to this assessment's causal factors or to its rivals', and chose a term each time. A second model, from a different family, checked each development a reading relied on, disputed 5 developments, and the five readings were made again without them. The record started the assessment at the average of the five final terms, taken in log-odds: 67.5 per cent, Likely (55–80%). Every reading and every check is kept in the first day's record.

The five final readings, in order: Likely, Likely, Likely, Likely, Likely. Their average, taken in log-odds, is 0.73, a stated confidence of 67.5 per cent. The readings were made by Claude Sonnet 4.6 (Anthropic) and checked by Gemini 4 Argon (Google).

Factors named above: M1 (Private deployments for large organizations), M2 (Agreements with governments, public bodies and defence suppliers), M3 (Partners that bring Cohere's models into private settings), M4 (Products designed to run under the customer's control), M5 (Signs of a turn toward the race for the most capable models), C1 (Private and sovereign offerings from larger rivals), C2 (Contracts won by rivals in Cohere's markets) and C3 (Strain at Cohere).

The developments shown

1. 1. First reported 9 October 2026, 2 independent publishers. Linked to a rival's causal factors.
   “Germany drops binding framework for sovereign AI partnership with Mistral and SAP” · sovereigntymonitor.eu · 9 October 2026 · Open the report · archived copy
   “Digital Ministry clarifies: No withdrawal from AI project with SAP and Mistral” · heise.de · 14 October 2026 · Open the report · archived copy
   First round: relied on by 1 reading as support for the assessment and 2 readings as counting against it. Second round: relied on by 1 reading as support for the assessment and 2 readings as counting against it.
   • Check of its use as counting against the assessment: Upheld by the second model, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The reports confirm that Germany has not withdrawn from its sovereign AI project with rival Mistral AI and SAP, and that the companies are continuing to develop public-sector AI technology for the German market. This ongoing rival effort in a target market directly aligns with factor C1 (sovereign offerings from rivals) and counts against Cohere's main assessment by supporting the rival competition scenario.

     • Report 2, quoted: “Digital Ministry clarifies: No withdrawal from AI project with SAP and Mistral”
     • Report 2, quoted: “noting the companies continued developing public-sector technology independently.”
   • Check of its use as support for the assessment: Upheld by the second model, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The reports state that Germany dropped a binding framework for a sovereign AI partnership with Mistral and SAP, with the Digital Ministry confirming there are no procurement or funding commitments. This weakens factor C2 (contracts won by rivals in Cohere's target markets, specifically Germany and rival Mistral AI), which directly reduces the likelihood of rival displacement (MXC) and thereby supports the main assessment.

     • Report 1, quoted: “Germany drops binding framework for sovereign AI partnership with Mistral and SAP”
     • Report 2, quoted: “Germany’s digital ministry said its letter of intent with SAP and Mistral was non-binding and involved no procurement or funding commitments”
2. 2. First reported 5 October 2026, 2 independent publishers. Linked to this assessment's own causal factors.
   “Cohere Launches North 2 With Redesigned Agent Harness and Memory” · unite.ai · 5 October 2026 · Open the report · archived copy
   “North 2: Cohere's latest platform update — with 15+ new features” · cohere.com · 5 October 2026 · Open the report · archived copy
   First round: relied on by 5 readings as support for the assessment. Second round: relied on by 5 readings as support for the assessment.
   • Check of its use as support for the assessment: Upheld by the second model, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The reports state that Cohere launched North 2 with enterprise governance controls and private, on-premises, and air-gapped deployment options. Releasing a new version of North designed for private deployment directly matches factor M4 and supports the assessment that Cohere is aiming its growth at secure, private AI for large organizations and governments.

     • Report 1, quoted: “Cohere Launches North 2 With Redesigned Agent Harness and Memory”
     • Report 2, quoted: “Cohere introduced North 2, adding agent orchestration, governance and cost controls, and flexible private, on-premises, and air-gapped deployment options for enterprise AI.”
3. 3. First reported 5 October 2026, 2 independent publishers. Linked to this assessment's own causal factors.
   “PwC & Cohere launch Canadian enterprise AI alliance” · itbrief.co.uk · 6 October 2026 · Open the report · archived copy
   “PwC and Cohere announce a global alliance to accelerate secure and trusted enterprise AI adoption” · pwc.com · 5 October 2026 · Open the report · archived copy
   First round: relied on by 5 readings as support for the assessment. Second round: relied on by 5 readings as support for the assessment.
   • Check of its use as support for the assessment: Upheld by the second model, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The reports confirm that PwC and Cohere announced a global alliance in October 2026 to help organizations deploy secure enterprise AI in private-cloud, on-premises, and air-gapped settings. This directly matches factor M3 (partnerships with consultancies that bring Cohere's models into private settings) and supports the main assessment over the listed alternatives.

     • Report 2, quoted: “PwC and Cohere announce a global alliance to accelerate secure and trusted enterprise AI adoption”
     • Report 2, quoted: “help organizations deploy enterprise AI, including private-cloud, on-premises, and air-gapped options for sensitive workloads.”
4. 4. First reported 14 October 2026, 1 independent publisher. Linked to a rival's causal factors.
   “Canadian AI startup Cohere buys Germany's Aleph Alpha to expand in Europe” · economictimes.indiatimes.com · 14 October 2026 · Open the report · archived copy
   First round: relied on by 4 readings as support for the assessment and 1 reading as counting against it. Removed from the second round, because the second model disputed it.
   • Check of its use as support for the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The report states that Cohere is buying Aleph Alpha and receiving a $600 million investment in its upcoming funding round, without reporting any new customer sales. This fits alternative MX2—which states that 'Cohere grows mainly through acquisitions and new capital rather than through sales: the merger with Aleph Alpha and new funding enlarge the company while its sales grow slowly'—at least as well as it fits the main assessment of organic growth through sales.

     Its ground: it fits one of the rival explanations as well as the assessment.
     • Report 1, quoted: “Canadian AI startup Cohere buys Germanys Aleph Alpha to expand in Europe”
     • Report 1, quoted: “Cohere agreed to acquire Germany’s Aleph Alpha to expand in regulated European markets, with Schwarz Group also set to invest $600 million in Cohere’s upcoming funding round.”
   • Check of its use as counting against the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: Factor C3 measures strain at Cohere, such as job cuts, missed revenue targets, or delays in its funding round or in its merger with Aleph Alpha. The report instead states that Cohere agreed to acquire Aleph Alpha to expand in regulated European markets and secured a $600 million investment commitment, which does not represent strain under C3 counting against the main assessment.

     Its ground: what the reports state is not a change in this causal factor.
     • Report 1, quoted: “Canadian AI startup Cohere buys Germanys Aleph Alpha to expand in Europe”
     • Report 1, quoted: “Cohere agreed to acquire Germany’s Aleph Alpha to expand in regulated European markets, with Schwarz Group also set to invest $600 million in Cohere’s upcoming funding round.”
5. 5. First reported 14 October 2026, 1 independent publisher. Linked to a rival's causal factors.
   “Trellix launches private cloud AI for security teams” · itbrief.com.au · 14 October 2026 · Open the report · archived copy
   First round: relied on by 1 reading as counting against it. Removed from the second round, because the second model disputed it.
   • Check of its use as counting against the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: Factor C1 covers private and sovereign offerings from 'larger rivals, among them the largest US AI and cloud companies and the French model maker Mistral AI, that let large organizations and governments run capable AI under their own control.' The report describes Trellix launching 'an on-premises AI system for security operations' for 'security teams,' which is a domain-specific cybersecurity tool rather than a general-purpose or frontier AI model offering from a larger AI or cloud rival competing with Cohere.

     Its ground: what the reports state is not a change in this causal factor.
     • Report 1, quoted: “Trellix launches private cloud AI for security teams”
     • Report 1, quoted: “Trellix launched Wise for Private Cloud, an on-premises AI system for security operations designed to run in private-cloud, hybrid, and air-gapped environments where sensitive data must remain within organizational control.”
6. 6. First reported 14 October 2026, 1 independent publisher. Linked to this assessment's own causal factors.
   “The Short Report: October 14, 2026” · researchmoneyinc.com · 14 October 2026 · Open the report · archived copy
   First round: relied on by 2 readings as support for the assessment. Removed from the second round, because the second model disputed it.
   • Check of its use as support for the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The report only states that Cohere announced an Ottawa office and launched North 2. It does not report any contracts or commitments by large organizations to run Cohere's models or its North platform under their own control, as required by factor M1. Furthermore, the North 2 product launch itself is already covered by earlier development [5], and opening an office would also fit equally with growth driven by new capital (MX2).

     Its ground: what the reports state is not a change in this causal factor.
     • Report 1, quoted: “Cohere announced its first Ottawa office and launched North 2, highlighting expansion of its Canadian operations and a new version of its work-automation and AI-agent system.”
7. 7. First reported 14 October 2026, 1 independent publisher. Linked to a rival's causal factors.
   “France carves its own path in AI race with Nvidia’s help” · euronews.com · 14 October 2026 · Open the report · archived copy
   First round: relied on by 1 reading as counting against it. Second round: relied on by 2 readings as counting against it.
   • Check of its use as counting against the assessment: Upheld by the second model, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The report describes France's sovereign AI push and a government deployment backed by Nvidia and involving rival model maker Mistral as an ecosystem partner. This represents sovereign AI activity in the European public sector by larger rivals and Mistral AI under factor C1, which supports alternative MXC and counts against Cohere's main growth assessment.

     • Report 1, quoted: “France carves its own path in AI race with Nvidia’s help”
     • Report 1, quoted: “The report describes France’s sovereign technology push and notes a government deployment involving a French digital workplace, while identifying Mistral as a partner in the country’s broader AI ecosystem.”
8. 8. First reported 13 October 2026, 1 independent publisher. Linked to a rival's causal factors.
   “SBS Launches Banking’s First Sovereign AI Platform” · einpresswire.com · 13 October 2026 · Open the report · archived copy
   First round: relied on by no reading. Second round: relied on by 1 reading as support for the assessment and 2 readings as counting against it.
   No reading relied on it, so it was not checked.
9. 9. First reported 30 September 2026, 1 independent publisher. Linked to this assessment's own causal factors.
   “Embed 5: Frontier Embedding Models for Enterprise” · cohere.com · 30 September 2026 · Open the report · archived copy
   First round: relied on by 3 readings as support for the assessment. Removed from the second round, because the second model disputed it.
   • Check of its use as support for the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

     Its reason, as the model wrote it: The report announces the product release of Embed 5 by Cohere, which the reading itself identifies as product evidence under factor M4. However, this development was linked only to M3 (partner agreements) and M1 (customer contracts and commitments for private deployments). Because the report states neither a new partner agreement nor a contract or commitment by a large organization to adopt the model, it is not a change in either of the shown factors.

     Its ground: what the reports state is not a change in this causal factor.
     • Report 1, quoted: “Cohere released Embed 5 for enterprise retrieval, with availability through its API and cloud platforms and support for private VPC or on-premises deployment.”
10. 10. First reported 25 September 2026, 1 independent publisher. Linked to this assessment's own causal factors.
    “Compass is coming to the cloud” · cohere.com · 25 September 2026 · Open the report · archived copy
    First round: relied on by 4 readings as support for the assessment. Removed from the second round, because the second model disputed it.
    • Check of its use as support for the assessment: Disputed by the second model, so the second round was read without it, run on Gemini 4 Argon (Google).

      Its reason, as the model wrote it: The report describes Cohere launching a private beta for its own managed Compass Cloud product while continuing to offer self-hosted options. This product launch does not involve an agreement with a partner to distribute Cohere's models in private settings (M3), nor does it report any contracts or commitments by large organizations to deploy Cohere's models under their own control (M1).

      Its ground: what the reports state is not a change in this causal factor.
      • Report 1, quoted: “Compass is coming to the cloud”
      • Report 1, quoted: “Cohere opened a private beta for managed Compass Cloud while retaining self-hosted deployments for privacy-constrained and regulated-industry customers.”

The readings, and their reasons

Readings made before this one

No starting point was read before this one, as the first day records.

Causes of uncertainty

• Cohere is a private company, so its revenue, and the share of it that comes from private deployments, is known only when Cohere chooses to say or when investor documents reach the press.
• Many of its announcements are memoranda of understanding or partnerships that commit no money, and these are weaker evidence than paid contracts.
• Once the merger with Aleph Alpha closes, growth from new customers and growth from the business it adds will be hard to tell apart.

What would change the judgment

• Cohere stating that most of its revenue now comes from use through shared public clouds or from individual developers.
• An agreement giving Cohere computing capacity on the scale of the largest AI labs.
• Growth that comes mainly from the businesses Cohere buys rather than from new customers.
• Cohere's revenue from large organizations and governments levelling off while larger rivals win contracts in its markets.

The causal factors, and which way each one cuts

What the assessment assumes, and what would break it

• MA1. Deployments under the customer's own control remain most of Cohere's business through 2027.
  If this fails, the assessment fails.

  Invalidated if Cohere states, in its newsroom, in remarks reported by Reuters, Bloomberg, the Financial Times, The Globe and Mail, The Logic or BetaKit, or in a prospectus it files with a securities regulator, that deployments in customers' own data centres, in private clouds or in systems Cohere runs for a single customer alone made up less than half of its revenue for a period ending after 30 September 2026. If no such figure is published before the end of 2027, the assumption is reviewed rather than invalidated.

  Source: Cohere newsroom, Reuters, AI firm Cohere doubles annualized revenue to $100 million on enterprise focus, 15 May 2025, as republished by Investing.com. The criterion’s fingerprint 0d8a8bc3…6ee7 was fixed on 15 October 2026 and is part of the record’s first day.
• MA2. Governments buy Cohere's private AI under paid contracts, not only agreements to explore it, beginning with the Government of Canada.
  If this fails, the assessment weakens.

  Invalidated if the Government of Canada's public list of its contracts over $10,000 (its proactive disclosure of contracts), searched for the vendor name Cohere, shows no new contract or contract amendment with Cohere awarded in the federal fiscal year from 1 April 2026 to 31 March 2027, when checked on or after 1 September 2027. If the Government of Canada stops publishing this disclosure, the assumption is reviewed rather than invalidated.

  Source: Government of Canada, Search Government Contracts over $10,000 (proactive disclosure), searched for Cohere. The criterion’s fingerprint 994e022c…342b was fixed on 15 October 2026 and is part of the record’s first day.
• MA3. Cohere stays out of the race to build the most capable general-purpose models and does not spend on computing at the scale of the largest AI labs.
  If this fails, the assessment weakens.

  Invalidated if, before the end of 2027, Cohere announces an agreement to buy, lease or build computing capacity for its own use that draws at least one gigawatt of power or is worth at least US$10 billion, as stated in its newsroom or reported by Reuters, Bloomberg or the Financial Times. Capacity that a partner operates to run customers' deployments, such as the Schwarz Group's STACKIT cloud, does not count.

  Source: Cohere newsroom, Cohere, Cohere and Aleph Alpha sign agreement to become the first transatlantic sovereign AI solution, 16 September 2026, which describes the partnership with STACKIT. The criterion’s fingerprint 32b28b48…b8dc was fixed on 15 October 2026 and is part of the record’s first day.
• MA4. Cohere completes its merger with the German AI company Aleph Alpha and gains a second home market in Germany.
  If this fails, the assessment weakens.

  Invalidated if Cohere or Aleph Alpha announces that the business combination agreement the two companies signed on 16 September 2026 has been terminated, or if Cohere has not announced, in its newsroom or a press release, that the merger is complete by 30 June 2027.

  Source: Cohere, Cohere and Aleph Alpha sign agreement to become the first transatlantic sovereign AI solution, 16 September 2026, Cohere newsroom. The criterion’s fingerprint aaf8d3dd…b827 was fixed on 15 October 2026 and is part of the record’s first day.

The rival explanations

Evidence moves the assessment only when it bears differently on the assessment than on at least one of these. Evidence that fits them equally carries no weight.

• MX1. Cohere grows mainly by competing on capability, training much larger general-purpose models and selling access to them through shared public clouds and tools for developers.
  No development was recorded on this assessment in the last 7 days.
  Favoured by:
  • Cohere announces computing capacity on the scale of the largest AI labs.
  • Cohere's new models are promoted as leading on general capability.
  • Cohere reports growth from developers and public cloud use rather than from private deployments.
  Ruled out by:
  • Cohere's new models remain sized to run on a few chips at the customer's site.
• MX2. Cohere grows mainly through acquisitions and new capital rather than through sales: the merger with Aleph Alpha and new funding enlarge the company while its sales grow slowly.
  No development was recorded on this assessment in the last 7 days.
  Favoured by:
  • Most of the growth in Cohere's revenue or staff comes from businesses it has bought.
  • Large funding announcements arrive without new customers of matching size.
  Ruled out by:
  • Cohere reports revenue growth from new customers well beyond what its acquisitions add.
• MXC. Larger rivals, among them the largest US AI and cloud companies and the French model maker Mistral AI, offer equivalent private and sovereign deployments and win the customers Cohere targets, so by the end of 2027 Cohere's sales of secure, private AI to large organizations and governments stop growing.
  This is the alternative assessment, stated word for word as the main assessment’s rival: the two cannot both be right.
  No development was recorded on this assessment in the last 7 days.
  Favoured by:
  • Government and large-organization contracts in Cohere's markets go to larger rivals that offer private deployment.
  • Large organizations that used Cohere move their AI work to larger rivals.
  • Cohere's revenue growth slows while rivals report wins with the same kinds of customer.
  Ruled out by:
  • Cohere wins contracts against larger rivals that offer private deployment, and its revenue from large organizations and governments keeps growing.

A development is recorded when the rule applies it to the assessment or sets it aside. Each count comes from the readings sealed with the changes, and a development set aside as fitting every alternative equally tells no alternative apart.

Evidence behind changes in the last 30 days

Each item is a report or data series cited by a sealed change on this assessment. The update loop reads headlines and summaries, and no summary is shown as the evidence.

No evidence has been cited in the last 30 days.

Evidence considered and set aside

Developments the update loop read in the last 30 days and set aside because they fitted this assessment and each of its rival explanations equally. Under the rule such evidence carries no weight, so none of it moved anything.

Nothing on this assessment was set aside in the last 30 days.

Changes to this assessment

No change to this assessment yet.

The rest of the judgment

• The condition it rests on
• The alternative

The question’s text on this page is as sealed on the record’s first day. Fingerprint 412510ac…302b. Fingerprint pending. Records are fingerprinted each day at 23:59 UTC.