First Time Founders: Is Cohere the Next AI Powerhouse?
Episode
57 min
Read time
2 min
Topics
Career Growth, Productivity, Startups
AI-Generated Summary
Key Takeaways
- ✓Enterprise-only positioning: Cohere deliberately excludes consumer products, targeting only medium-to-large enterprises with deployment models that keep customer data private and inaccessible to Cohere itself. This produces SaaS-like margins rather than the losses consumer AI companies absorb per user, creating a fundamentally different and more sustainable financial structure for eventual public markets.
- ✓Foundational model barriers to entry: Roughly 10 companies worldwide can build large language models because the process resembles rocket engineering — requiring massive compute clusters, enormous curated datasets, human annotation teams, and hundreds of specialized engineers working in tight coordination. This concentration exists across only four countries: the US, Canada, China, and France.
- ✓Three-stage training pipeline: Modern LLMs are built through sequential data layers — first training on the entire open web, then fine-tuning on human-generated chat dialogues with rated responses, then running reinforcement learning on synthetic data the model generates itself. Chat fine-tuning specifically was the underestimated breakthrough that made models accessible to non-technical users after 2022.
- ✓AI's realistic productivity ceiling: Current transformer-based AI can automate roughly 20–30% of desk-based knowledge work across all organizational levels, not just entry-level roles. It cannot replace strategic thinking, cultural interpretation, or interpersonal coordination. Framing AI as a full job replacement rather than a productivity multiplier misrepresents what the technology actually does at this stage.
- ✓Career advice under technological uncertainty: Rather than chasing predicted high-demand roles — which forecasters consistently get wrong — young people should optimize for personal curiosity and genuine interest. Intrinsic motivation produces higher performance and adaptability than strategically chosen career paths, particularly in chaotic technological transitions where the landscape shifts faster than any prediction model can track.
What It Covers
Nick Frost, cofounder of Cohere — a $7 billion enterprise AI company founded in 2019 by three former Google engineers — explains why only 10 companies globally can build foundational models, how Cohere differs from OpenAI and Anthropic, and why AGI remains a distraction from AI's real economic impact.
Key Questions Answered
- •Enterprise-only positioning: Cohere deliberately excludes consumer products, targeting only medium-to-large enterprises with deployment models that keep customer data private and inaccessible to Cohere itself. This produces SaaS-like margins rather than the losses consumer AI companies absorb per user, creating a fundamentally different and more sustainable financial structure for eventual public markets.
- •Foundational model barriers to entry: Roughly 10 companies worldwide can build large language models because the process resembles rocket engineering — requiring massive compute clusters, enormous curated datasets, human annotation teams, and hundreds of specialized engineers working in tight coordination. This concentration exists across only four countries: the US, Canada, China, and France.
- •Three-stage training pipeline: Modern LLMs are built through sequential data layers — first training on the entire open web, then fine-tuning on human-generated chat dialogues with rated responses, then running reinforcement learning on synthetic data the model generates itself. Chat fine-tuning specifically was the underestimated breakthrough that made models accessible to non-technical users after 2022.
- •AI's realistic productivity ceiling: Current transformer-based AI can automate roughly 20–30% of desk-based knowledge work across all organizational levels, not just entry-level roles. It cannot replace strategic thinking, cultural interpretation, or interpersonal coordination. Framing AI as a full job replacement rather than a productivity multiplier misrepresents what the technology actually does at this stage.
- •Career advice under technological uncertainty: Rather than chasing predicted high-demand roles — which forecasters consistently get wrong — young people should optimize for personal curiosity and genuine interest. Intrinsic motivation produces higher performance and adaptability than strategically chosen career paths, particularly in chaotic technological transitions where the landscape shifts faster than any prediction model can track.
Notable Moment
Frost describes a 500-year-old Yiddish folktale about a rabbi who animates a clay man and instructs it to fetch fish — returning to find the river emptied and his house flooded. He uses this to argue that the core anxieties around literal AI interpretation predate computers by centuries.
Episode Transcript
Support for today's show comes from Darktrace. Darktrace is the cybersecurity defenders deserve and the one they need to defend beyond. Darktrace is AI cybersecurity that can stop novel threats before they become breaches across email, clouds, networks, and more. With the power to see across your entire attack surface, cyber defenders, including IT decision makers, CISOs, and cybersecurity professionals now have the ability to stop zero days before day zero. The world needs defenders. Defenders need Darktrace. Visit darktracetrace.com/defenders for more information. Welcome to First Time Founders. I'm Ed Elson. Artificial intelligence has become one of the most heavily funded sectors in the world. More than 30 startups have raised over $100,000,000 this year alone. And as AI becomes more embedded in how the world operates, a handful of firms have emerged as the key players behind that transformation. Among them is a company building the kind of AI most people don't see. That is the AI that is powering the systems that run businesses and governments. Founded in 2019 by three former Google engineers, this company has focused squarely on the enterprise market, developing large language models for clients like Dell, SAP, and Salesforce. It even recently signed a deal with Canada's government to bring its technology into public operations. Now valued at nearly $7,000,000,000, it has earned a place alongside giants like OpenAI and Anthropic, helping define what the next era of AI will actually look like. This is my conversation with Nick Frost, cofounder of Cohere. Alright. Nick Frost, good to have you on the program. Thanks for having me. So for those who don't know, what Cohere is, I think we should probably just start there. What is Cohere? What does Cohere do? What are you guys building in AI? So we're a foundational model company, and we are uniquely and singularly focused on the enterprise. So there's there's about 10 companies in the world that can make foundational models. So, foundational models, the the large language models that are largely these days synonymous with AI. If somebody's talking about AI, they're they're probably talking about large language models. There's about 10 companies in the world that can make them. We are unique amongst them in our singular focus on the enterprise. So we make large language models that are good at the stuff that enterprises need them to be good at. We make them, easy to deploy and efficient to deploy for enterprises. We deploy them securely and privately so that we can't see the data that our customers are passing into the model that allows them to access the truly useful data out there. And we make them easy to work with via an agentic platform. So we do kind of the whole thing in order to get AI to work at work. So these foundational models, I think most people who are interested in tech kind of know what they are, but just at a very basic level, the foundational models are …
Get the full transcript (10,239 words) + summary by email — free
One-time email with the complete transcript and AI summary of this episode. No account needed.
One email, no spam. We’ll also show you what SignalCast does.
You just read a 3-minute summary of a 54-minute episode.
Get The Prof G Pod summarized like this every Monday — plus up to 2 more podcasts, free.
Pick Your Podcasts — FreeKeep Reading
More from The Prof G Pod
What Actually Predicts Divorce, and What Women Really Want in a Partner — ft. James Sexton
Sep 9 · 34 min
a16z Podcast
Building the Physical AI Stack | Travis Kalanick on TBPN
Jul 23
More from The Prof G Pod
China Decode: What's at Stake Ahead of Trump and Xi's September Meeting, and China's Biotech Surge
Sep 8 · 48 min
Eye on AI
#321 Nick Frosst: Why Cohere Is Betting on Enterprise AI, Not AGI
Feb 17
More from The Prof G Pod
We summarize every new episode. Want them in your inbox?
What Actually Predicts Divorce, and What Women Really Want in a Partner — ft. James Sexton
China Decode: What's at Stake Ahead of Trump and Xi's September Meeting, and China's Biotech Surge
No Mercy / No Malice: Little Tobacco Moment
The Week: When AI Escapes, and Wars Don't End
The AI Revolution Isn’t What Silicon Valley Promised — with Josh Tyrangiel
Similar Episodes
Related episodes from other podcasts
a16z Podcast
Jul 23
Building the Physical AI Stack | Travis Kalanick on TBPN
Eye on AI
Feb 17
#321 Nick Frosst: Why Cohere Is Betting on Enterprise AI, Not AGI
20VC (20 Minute VC)
Sep 1
20VC: Cohere Founder on How Cohere Compete with OpenAI and Anthropic $BNs | Why Counties Should Fund Their Own Models & the Need for Model Sovereignty | How Sam Altman Has Done a Disservice to AI with Nick Frosst
20VC (20 Minute VC)
Aug 10
20VC: Will OpenRouter Sell for $10BN to Stripe? | Why Chinese Open Models Are Beating America—and What Happens Next | Why Enterprises Are More Fearful of Anthropic and OpenAI Than China | Is the Routing Layer Becoming a Commodity with Alex Atallah
No Priors: Artificial Intelligence | Technology | Startups
Jul 31
Building an Autonomous Enterprise for Real-World Services with Netic Founder Melisa Tokmak
Explore Related Topics
This podcast is featured in Best Business Podcasts (2026) — ranked and reviewed with AI summaries.
Read this week's Startups & Product Podcast Insights — cross-podcast analysis updated weekly.
You're clearly into The Prof G Pod.
Every Monday, we deliver AI summaries of the latest episodes from The Prof G Pod and 192+ other podcasts. Free for one show.
Start My Monday DigestNo credit card · Unsubscribe anytime