May 24, 2024 · 18m · news
Sam Altman, Arthur Mensch and more discuss:Which Startups Are Threatened vs Enabled by OpenAI?|E1156 · 20VC with Harry Stebbings
compilation · excluded from per-person scoring
Tom Hulme · 4m spoken Harry Stebbings · 3m spoken Des Traynor · 1m spoken Sam Altman · 1m spoken Miles Grimshaw · 1m spoken Sarah Tavel · 1m spoken Arthur Mensch · 1m spoken Tom Blomfield · 58s spoken Tomasz Tunguz · 34s spoken Emad Mostaque · 32s spoken Brad Lightcap · 25s spoken
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
This video features a panel of prominent AI founders, venture capitalists, and software executives who debate the rapid commoditization of foundational AI models and outline strategic pathways for startups to build highly defensible, value-generating applications without being displaced by technology giants.
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 4.7 Guest teaching 4.8 Guest disagreement 3.0 Harry pushing back 3.2
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 11:35 Sam Altman's steamroll warning
Sam Altman bluntly asserts that OpenAI will steamroll any startup building on the assumption that underlying base models will stop improving.
Hardest push from Harry ▶ 7:12 Harry pressing on OpenAI's $90B valuationHarry directly puts guests on the spot by asking whether they would personally invest in OpenAI at a $90 billion valuation.
Biggest teaching moment ▶ 2:33 Tom Hume's power station mental modelTom Hume reframes foundation model investing through a power station analogy, explaining how rapidly depreciating capital assets destroy investor returns.
Harry holds his own ▶ 9:12 Harry citing Web2 cloud vs app market cap breakdownHarry demonstrates strong venture knowledge by introducing historical Web2 data on value capture split between infrastructure and application layers.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| The Model Commoditization Debate & Mistral's Strategy | 2 | 3 | 1 | 1 | Harry acts as a polite moderator inviting Arthur Mensch to respond to model commoditization. Sam Altman and Arthur Mensch calmly outline their strategic perspectives on base model advancement versus application layer efficiency. | |
| Foundation Models as a Temporary Power Station Asset | 5 | 6 | 2 | 4 | Harry presses Tom Hume on whether foundation models can yield returns given massive capital needs and dilution. Tom Hume educates with a power station depreciation mental model and Clay Christensen's framework on sustaining innovation. | |
| Intercom's Lens on Model Commoditization and OpenAI's Valuation | 6 | 5 | 3 | 5 | Harry directly challenges panel guests on whether they would invest in OpenAI at a $90 billion valuation. Des Traynor and Tom Hume explain why high valuations are risky due to fast commoditization and cloud provider competition. | |
| The Oligopoly of Foundation Model Companies | 5 | 5 | 5 | 4 | Harry interrupts Emad Mostaque to ask specifically about Anthropic and probes OpenAI leadership on founder risk. Sam Altman aggressively warns that startups assuming static models will be steamrolled by OpenAI's rapid progress. | |
| Thick Wrappers and Deep Industry Integration | 4 | 5 | 3 | 2 | Harry guides the conversation around thin vs thick wrappers across guests. Des Traynor uses a train tracks metaphor while Tom Blomfield argues deep domain workflows are safe from general model providers. | |
| Copilots as Incumbent Strategies vs. Selling the Work | 6 | 5 | 4 | 3 | Harry shares his own VC experience on copilots as entry wedges to prompt Myles Grimshaw, who counters that copilots favor incumbents. Sarah Tavel completes the dynamic by explaining the shift toward selling completed work outcomes over seats. |
Statements from this episode (19)
Sam Altman: AI model market will consolidate into dozens of large providers
“I think eventually it will shake out. There will be a small number of providers, just dozens, something like that, doing models at Big scale.”
Sam Altman: Long-term AI value lies in personalization, not base models
“The long-term differentiation will not be, I don't think the base model, like that's just, you know, intelligence is just like some emergent property of matter or something. The long-term differentiation will be the model that's most personalized to you, that …”
Tom Hulme: Meta will own 14% of global Nvidia H100s by year-end
“He'll have 350,000 H-one hundreds by the end of this year. That is 14% of the world's H-one hundreds, and he's going to open source the result.”
Tom Hulme: Google Ventures avoids foundation models to focus on AI apps
“Now we have made investments in Gen AI, but more in infrastructure, more in the application layer, more in the sort of picks and shovels to support, but we've not thus far invested in foundation models.”
Tom Hulme: Generative AI will not cause internet-level creative destruction
“It's actually going to get sprinkled across all businesses to lower costs in call centers or to improve the product in personalization. It's not going to Have a creative destruction effect like the internet did on many industries.”
Des Traynor: Intercom passes substantial software value directly to OpenAI
“Right now, a lot of value is going straight into the infra, like as in we're handing it all out the back door to open AI, and we actually torture test all of the LLMs.”
Des Traynor: OpenAI valuation will pass $90B, but Amazon poses risks
“I don't think I would, and the reason, whilst I think they'll pass 90, the areas I'd be wary of is Amazon.”
Tomasz Tunguz: Web2 cloud infra and top 100 apps share equal market caps
“So in web two, if you take the top three clouds and you look at their market cap, so AWS, GCP, and Azure, It's about a 2.1 trillion dollar market cap just for the cloud businesses. And then if you take the top 100 publicly traded cloud companies, both on B to …”
Tomasz Tunguz: Investor success odds are significantly higher at AI application layer
“If the analogy holds, as an investor, the odds of success are going to be significantly higher at the application layer because the diversity of needs there is greater.”
Emad Mostaque: Only 5 or 6 foundation model companies will survive
“I think that there's only going to be five or six foundation model companies in the world in three years, five years. I think it's going to be us, Nvidia, Google, Microsoft, OpenAI, and Meta, and Apple probably are the ones that train these models.”
Emad Mostaque: Google spends $20 billion annually on artificial intelligence
“Google spend twenty billion dollars a year on AI.”
Emad Mostaque: Google DeepMind's annual salary budget is $1.2 billion
“DeepMind's salary budget is 1.2 billion a year.”
Sam Altman: OpenAI will steamroll startups built on static model assumptions
“It would seem to me that 95% of the world should be betting on the latter category, but a lot of the startups have been built in the former category. When we just do our fundamental job, we're going to steamroll you.”
Brad Lightcap: 100x model improvement is the key test for startup defensibility
“Ask the company whether a hundred X improvement in the model is something they're excited about. It's actually, we can tell pretty well because we know the companies that come to us saying, we want the next model. When is it coming out? When is it coming out? …”
Des Traynor: OpenAI will never build deep vertical industry integrations
“OpenAI is never going to put like five of their engineers going hard on wealth management and like, you know, banking integrations.”
Tom Blomfield: AI applications are 80-90% traditional software and 10% AI
“Most people building application layer stuff in an AI say it's 80 to 90% traditional software with 10% AI.”
Grimshaw: Building AI copilots is an incumbent strategy, not a startup opportunity
“I think Copilot is an incumbent's strategy. Incumbents own distribution, they own data, they own the UX, and they own a business model that all aligns to a Copilot.”
Sarah Tavel: AI enables software startups to sell completed work outcomes
“What AI enables is actually a very different unit of work that you sell, which is doing the work. And so you're almost a software company that looks like a services business. That is able to sell like the full work product that the outcome, as opposed to selli…”
Sarah Tavel: Non-seat-based AI models will disrupt traditional SaaS incumbents
“And this is very disruptive to incumbents because incumbents are used to thinking about selling per seat and pricing per seat based on the cost of the headcount. But if instead you're selling something that doesn't require a seat, that is like a very disruptiv…”