Investors should focus on the Foundational Model layer of the AI stack, as this high-barrier oligopoly acts as essential national infrastructure with long-term government support. Cohere represents a high-conviction play in the enterprise sector, offering SaaS-like margins by deploying secure, private models for regulated industries rather than volatile consumer markets. You can gain exposure to this enterprise AI adoption through key integration partners and clients like Salesforce (CRM), SAP (SAP), and Dell (DELL). Monitor NVIDIA (NVDA) as the primary provider of the massive compute required for these "rocket ship" foundational models to maintain their competitive moats. For long-term growth, keep Cohere on an IPO Watch as they aim to become a generational public company by automating high-value business workflows.

Cohere

• Cohere is a foundational model company valued at nearly $7 billion, founded by former Google engineers (including a co-author of the seminal "Attention is All You Need" paper). • The company distinguishes itself from giants like OpenAI and Anthropic by being "singularly focused on the enterprise." • Unlike consumer-facing AI, Cohere does not offer a $20/month subscription for individuals. • They specialize in Large Language Models (LLMs) tailored for business efficiency, security, and private data handling. • Key Clients & Partners: They have established deals with Dell, SAP, Salesforce, and the Canadian government. • Product Strategy: Their platform, North, is an agentic framework that allows businesses to automate complex workflows, such as cross-referencing emails with Salesforce data or analyzing private data rooms.

Takeaways

• Enterprise vs. Consumer: Investors should note that Cohere is positioning itself as the "secure" alternative for regulated industries. While OpenAI captures the headlines, Cohere is capturing the "boring" but lucrative back-end infrastructure of global corporations. • SaaS-like Margins: Unlike consumer AI models that often lose money on every query due to high compute costs, Cohere’s business model mirrors SaaS (Software as a Service) margins because they deploy models directly into a client's secure environment. • IPO Watch: While no specific timeline was given, the co-founder confirmed that going public is the intended path to becoming a "generational company."

Foundational AI Models (The "Rocket" Business)

• There are only about 10 companies globally capable of building foundational models (the "backbone" of AI). • High Barriers to Entry: Building these models is compared to "building a rocket." It requires: • Massive amounts of compute (chips). • Enormous data sets and human annotators. • Highly specialized engineering talent working in tight unison. • The Four-Nation Race: Currently, only four countries—the US, China, France, and Canada—host companies capable of producing this level of technology.

Takeaways

• Concentrated Power: The "Foundational Model" layer of the AI stack is an oligopoly. For investors, this suggests that while there are thousands of AI startups, the real long-term value may reside in the few companies that own the underlying "infrastructure." • Infrastructure Play: AI is increasingly viewed as national infrastructure, similar to nuclear power plants or roads. This suggests long-term government support and subsidies for domestic champions like Cohere (Canada) or Mistral (France).

AI Investment Themes & Sectors

The "AGI" Skepticism

• Cohere’s leadership expresses a "heretical" view: Transformers (the current AI architecture) will likely not lead to Artificial General Intelligence (AGI). • They argue that the current obsession with "digital gods" or human-like intelligence is a narrative device rather than a scientific reality.

Labor Market Transformation

• Augmentation, not Replacement: The discussion suggests AI will automate 20-30% of tasks for most desk jobs rather than 100% of any single job. • Historical Parallel: The current moment is compared to the Industrial Revolution or the transition from typists to computer users. It will be "chaotic" but ultimately value-accretive for the economy.

Takeaways

• Focus on Utility over Hype: Actionable insights suggest looking for companies using AI to solve specific, "un-enjoyable" business tasks (data entry, document analysis, scheduling) rather than companies promising "human-level" AI. • Policy Risk: Wealth inequality is cited as a major risk factor. Investors should monitor potential government policies regarding income distribution or AI regulation, as these could impact the profitability of AI "owners."

Mentioned Tickers & Entities

• Salesforce (CRM): Mentioned as a key integration partner and client for Cohere’s enterprise AI. • SAP (SAP): Identified as a major enterprise client using Cohere models. • Dell (DELL): Identified as a partner/client in the enterprise AI space. • Microsoft (MSFT): Mentioned in the context of its massive contract and partnership with OpenAI. • NVIDIA (NVDA): Referenced indirectly via the "massive compute" and "chips" required to build foundational models.