May 12, 2023, 4:34 PM ETInterview, Fireside Chat

Clem Delangue: The Ultimate Guide to Investing in AI; Elon's Threat to Sue OpenAI | E1013

Hugging Face was named after the "hugging face" emoji to serve as a unique public market ticker alternative, a decision that became central to the brand's identity due to massive community adoption.
The company was founded seven years ago by Clem and co-founders Julia and Thomas driven by a desire to work together and early excitement regarding AI's potential as a new technological paradigm.
The initial product concept was a "Tamagotchi AI" or entertainment-focused chatbot, which ran for nearly three years and facilitated billions of user messages before pivoting.
A strategic pivot occurred after the community demonstrated significant traction with the underlying open-source platform technology, shifting the company from a consumer entertainment focus to an AI infrastructure platform.

Clem argues that current VC and mainstream excitement is a "catch-up" phase to actual AI usage, which has been massive for years in products like Google search, Facebook feeds, and Zoom background removal.
Recent breakthroughs by OpenAI (ChatGPT) and hardware optimizations (GPU availability, quantization, distillation) are viewed as the final pieces required to move AI from niche usage to mainstream scale.
The rapid progress in AI is attributed primarily to open science and open-source collaboration, with researchers sharing foundational papers (e.g., "Attention Is All You Need," "Latent Diffusion") that allow the industry to build upon existing work.
Content access and training data legal status are identified as significant future obstacles, with upcoming regulatory clarity expected regarding fair use and how companies must respect content creator rights.

The interviewee rejects the notion that AI startups must be located in Silicon Valley, citing Hugging Face's distributed model with teams in Paris, New York, and San Francisco.
While Silicon Valley retains high energy, AI talent and scientific breakthroughs are heavily decentralized; for instance, 10 of the 13 authors of the Llama model were based in Meta's Paris lab.
Founders are advised to prioritize personal happiness and location preference over geography, with the argument that founders can build successful companies from any location if they are content.
Hugging Face operates with a decentralized structure to leverage global talent, maintaining a "globalized talent network" rather than a centralized hub.

The industry is divided between the "one model to rule them all" approach (concentrated capability, API reliance) and the "many models" approach (distributed, companies building and optimizing their own models).
Relying on a single external API (e.g., OpenAI) offers short-term speed but creates long-term risks of high costs, lack of differentiation, and an inability to optimize models for specific use cases.
AI-native startups are advised to train and optimize their own models to achieve 10x to 100x performance improvements over incumbents, as large organizations struggle to adapt to the scientific architecture required for custom model building.
Incumbents risk capturing 90% of the gains if they simply wrap existing APIs into existing products (e.g., Microsoft, Adobe), whereas true value creation comes from startups that fundamentally re-architect technology using custom AI.
Enterprises may initially prefer bundled, easy-to-use solutions, creating an opportunity for startups to disrupt these "deep cut bonds" by offering superior, tailored AI-native solutions.

Hugging Face utilizes a freemium model where approximately 15,000 companies use the platform, with 3,000 paying for premium features such as single sign-on, premium support, and accelerated compute (GPUs).
Revenue prioritization focuses on adoption and usage (network effects) rather than immediate monetization, operating on the assumption that usage will eventually convert to revenue.
Pricing models vary and have not yet been fully optimized, as the platform views monetization as a progressive learning process to adapt to a rapidly evolving technology landscape.
The interviewee views the primary goal as building the most impactful organization in AI, with being the "biggest company" viewed as a potential side effect rather than a core objective.

Clem enforces a rule against speaking with external investors between fundraising rounds to maintain focus, only engaging during active rounds to ensure serious intent.
The interviewee pushes back against the practice of sending term sheets without prior relationship building, arguing that investors invest in "lines, not dots," and that authentic trust requires a longer-term relationship.
Fundraising difficulty is dictated more by traction and momentum at a specific moment than by the specific round stage (e.g., a Series C can be easier than a Pre-seed depending on market conditions).
Venture capitalists are criticized for shifting focus from their core value of financial capitalization to operational interference, sometimes causing startups to build companies for investors rather than for customers.

The single biggest challenge for AI-first startups is hiring hybrid talent capable of both scientific research and engineering, as the supply of experts who can build new architectures is extremely limited (estimated at 50–100 people globally).
High salaries driven by massive early-stage funding rounds ($100M–$200M raises) have created intense competition for top-tier AI talent.
The existential market risk for Hugging Face is the failure of the broader AI sector to deliver value, which is why the company prioritizes community support and open-source contribution to ensure ecosystem success.
Clem dismisses the "AGI taking over the world" narrative as sci-fi fear-mongering that distracts from immediate, solvable challenges like model bias, misinformation, and regulatory compliance.
The interviewee advises founders to accept that entrepreneurial struggles do not diminish as companies grow; each stage presents new challenges, and the joy should be found in the building process itself rather than the eventual exit.
Richard Sucher, former Chief Scientist at Salesforce and current founder of You.com, is identified as the most impactful angel investor due to his deep scientific background and business acumen.