May 12, 2023 · 58m · news
Clem Delangue: The Ultimate Guide to Investing in AI; Elon's Threat to Sue OpenAI | E1013 · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, Hugging Face co-founder and CEO Clem Delangue discusses the strategic advantages of the open-source AI ecosystem over monolithic closed models, while sharing his contrarian philosophies on startup execution, pragmatic AI regulation, and unorthodox fundraising strategies.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 21.9% of the talking time here. How this is scored →
Harry as informed peer 4.9 Guest teaching 4.3 Guest disagreement 3.0 Harry pushing back 4.3
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 40:54 Clem defends refusing VC meetings between rounds
Clem firmly rejects Harry's assertion that investors need long-term relationship building, insisting that taking meetings between rounds is a waste of founder focus.
Hardest push from Harry ▶ 40:07 Harry rejects Clem's fundraising methodologyHarry directly tells Clem that his fundraising approach is wrong, arguing that investors invest in lines rather than dots and calling Clem's method a shotgun marriage.
Biggest teaching moment ▶ 5:12 Clem re-educates Harry on VC hype vs usageClem corrects Harry's framing of AI hype, explaining that VC interest is actually a delayed catch-up to existing massive platform usage across Google, Facebook, and Zoom.
Harry holds his own ▶ 26:58 Harry challenges startup value capture using Alex Rampell thesisHarry demonstrates strong market expertise by quoting Alex Rampell to argue that fast-moving incumbents with distribution will capture 90% of the value in the AI boom.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Hook: Why the Startup Journey Never Gets Easier | 2 | 2 | 1 | 1 | Harry introduces Clem and asks lightweight background questions regarding the origin of the name Hugging Face and its initial founding story. Clem explains the playful emoji background and their pivot from an AI Tamagotchi to an open-source platform. | |
| AI Hype Cycles vs. Long-Term Technological Progress | 3 | 4 | 2 | 3 | Harry asks if current AI interest is driven by investor hype versus underlying technology progress. Clem reframes the question, clarifying that VC interest is actually a lagging catch-up to existing massive usage across tech incumbents. | |
| Decentralized Global Talent vs. Silicon Valley | 4 | 4 | 3 | 4 | Harry presses Clem on whether AI startups must move to Silicon Valley or if talent is globalized. Clem pushes back on Valley centrality by pointing out that most authors of Meta's LLaMA model are based in Paris, emphasizing founder happiness over geographic relocation. | |
| Monolithic Closed Models vs. Open Source Ecosystems | 6 | 4 | 2 | 6 | Harry uses Lee Fixel's framework comparing single monolithic models to open multi-model ecosystems and pushes back on Clem by citing slow enterprise procurement habits that favor bundled solutions. Clem argues that AI-native startups will disrupt slow incumbents by building specialized models. | |
| Legal Barriers, Training Data, and Regulations in AI | 5 | 3 | 2 | 3 | Harry cites Yann LeCun's quote on training data legality and asks about content providers like Reddit charging for data access. Clem agrees data legality is a shared challenge across open and closed AI, showcasing BigCode's opt-out dataset initiative. | |
| Hugging Face’s Freemium Business Model and Revenue Strategy | 5 | 5 | 3 | 5 | Harry asks direct questions about Hugging Face's monetization model and candidly asks if Clem gets annoyed by revenue questions. Clem explains their freemium strategy and argues that platform usage represents delayed revenue. | |
| Incumbents with Distribution vs. AI-First Startups | 7 | 5 | 4 | 6 | Harry challenges Clem's startup optimism by citing Alex Rampell's thesis that fast-moving incumbents with distribution (like Microsoft and Adobe) will capture 90 percent of AI value. Clem counters that incumbents struggle with true AI-first R&D paradigms. | |
| Challenges and Financial Realities of AI-First Startups | 5 | 5 | 2 | 4 | Harry probes into the financial realities of AI startups and questions the necessity of massive 100-200 million dollar early rounds. Clem confirms AI startup costs are higher but notes that diminishing returns on compute scaling may alter this fundraising dynamic. | |
| AI Safety Regulations vs. the Hype of AGI and Sci-Fi Threats | 5 | 5 | 5 | 4 | Harry quotes Elon Musk's call for preemptive AI regulation. Clem strongly rejects Musk's framing, dismissing sci-fi AGI doom narratives in favor of regulating practical, immediate issues like bias and misinformation. | |
| Clem's Unorthodox Fundraising Rules and Venture Capitalism | 8 | 6 | 7 | 9 | Harry aggressively challenges Clem's strict rule of refusing to meet VCs outside active fundraising rounds, calling it a flawed 'shotgun marriage' approach. Clem holds his ground firmly, arguing that out-of-round VC meetings are an inefficient distraction for founders. | |
| Startup Struggles, Enjoying the Journey, and Fundraising Stages | 4 | 4 | 2 | 2 | Clem reflects on the reality that startup challenges never get easier, only different. Harry transitions into quickfire questions covering AI market risks, favorite angels, and hiring bottlenecks. |
Statements from this episode (21)
Hugging Face Processed Billions of Chatbot Messages Before Open-Source Pivot
“We actually did that for almost three years. We raised our pre-seed and seed rounds on, on this idea. We got a product out with a couple of billion messages exchanged between users and this Tamagotchi AI.”
Delangue: VC Interest in AI Lags Behind Years of Massive Usage
“The VC and the mainstream interest is kind of like a catch up on the reality. Because if you look at usage of AI, it's been massive and it's been growing massively for the past, you know, three years, right? Even before, before ChatGPT, even before like, you k…”
Delangue: Open Source Accelerated AI Progress by 30 to 50 Years
“Without open science, without open source, without Google sharing their attention is all you need paper, sharing their birth paper the latent diffusion paper. Maybe we would be 3040, 50 years away from where we are today.”
Delangue: 10 of 13 Meta LLaMA Authors Are Based in Paris
“I don't have the exact number, but I think 10 out of the 13 authors of Lama are actually based in, in Paris, right? In, in the Meta AI, lab that, that they have there that is huge.”
Delangue: No Single AI Model Will Dominate Because AI Resembles Codebases
“In the simplistic terms, the model or like AI is like a code base, right? It's a bit different, but at the end of the day, it's a code base, right? And so it's kind of silly to say, oh, this code base is better than this code base, or there's going to be one c…”
Delangue: Using AI APIs Is Like Building Websites on Squarespace
“The analogy that I sometimes like is that on the early days of the web, you could use kind of like something that would Create a website for you, right? You could use the equivalent of a Squarespace or for weeks, and that would make you feel good, right? Becau…”
Delangue: AI Industry Will Gain Fair Use Regulatory Clarity in 2023
“And it's gonna be good this year, I think, because we're going to start to have more legal clarity about, you know, how do we consider fair use what are the regulators expecting from AI companies to respect in terms of rules.”
Delangue: Hugging Face Was First to Train AI on Opt-Out Datasets
“I think we were the first organization to train on a fully opted out data sets where developers could just remove themselves from the training.”
Hugging Face Has 15,000 Corporate Users and 3,000 Paying Customers
“We have 15,000 companies using us now. And then a smaller subset of companies are actually paying us, right? And for us, it's 3000 companies.”
Meta, Bloomberg, and Grammarly Are Paying Enterprise Customers of Hugging Face
“3000 companies are paying us for that, including Meta, including Bloomberg, including Grammarly and companies like that.”
Delangue: Hugging Face's Monetization Model Will Shift Within Five Years
“Probably the way we make money today is not going to be the way we make money in three years or in five years.”
Delangue: Incumbents Struggle With AI Due to Long-Term Research Requirements
“Because it's a completely different way to build technology, right? It's a way where you have to have scientists, for example, working for six months on a new architecture, on a new model before releasing it. So it's a different paradigm of how you build softw…”
Delangue: AI Startups Cost More to Build Than Traditional Software Startups
“It does, a little bit. A bit similarly, in my opinion, to how, you know, like you would build like an internet company or like a software company, 20 years ago. Because compute is kind of like more expensive for AI than it is for traditional software. Because …”
Delangue: The ROI on Training Larger AI Models Is Declining
“We starting to realize that more compute for models is not necessarily The right thing, or at least that it's not enough and the return on investment on training larger and larger models is starting to go down.”
Delangue: AI Is Far From Becoming Autonomous or Humanity-Destroying
“I believe at the same time, we are very far from a world where AI is autonomous. And has conscience and is taking over the world and destroying humanity. I think this is more fear that is very sci-fi driven that we're very far from.”
Delangue: Hugging Face Refuses to Speak With External VCs Between Rounds
“One of these rules is that I don't talk to any investors external investors in between rounds, right? I'm making an exception for you today because it's a podcast, but Otherwise, I don't talk to anybody in between rounds because I feel like a lot of time it's …”
Delangue: 95% of a VC's Value Is Financial Capitalization and Support
“Investors are first and foremost investors, right? Meaning that their main value add is to deal around To help you on financial matters. So for example, when SVB go down, right? And then to help you to always capitalize the company the right way. So they're, i…”
Delangue: VCs Should Not Act Like Operators or CEOs
“They sometimes act almost as CEO or, like, operators for companies. Which in, in my opinion is, is not really their job. And worse than that, I feel like sometimes entrepreneurs are building companies for investors and investors are behaving like entrepreneurs…”
Delangue: Fundraising Difficulty Depends on Momentum, Not Company Stage
“Something I didn't expect is that the way you raise Isn't so much dictated by your stage and your rounds. That's more dictated by the situation that you're in, in terms of traction, in terms of momentum, in terms of like achievements, right?”
Delangue: Every Company Will Eventually Build and Run Proprietary AI Models
“I think the biggest thing is, in my opinion, all companies will have their own AI models. Like all companies will have their chat GPT or their GPT-IV.”
Delangue: Only 50 to 100 People Globally Can Train State-of-the-Art AI
“I would say, you know, machine learning engineer, and by machine learning engineer, I mean someone who's really building a new architecture for AI models and able to train state of the art models. They are just in my opinion, a few people in the world who has …”