Overview of 20VC
This episode of 20VC with Harry Stebbings, Rory O’Driscoll, and Jason Lemkin is a fast-moving debate on how AI is reshaping tech valuations, product categories, talent retention, and executive incentives. The core theme: AI is no longer just a feature — for many companies, it’s a platform shift that can either compress growth or make the product obsolete.
The conversation focuses on Canva’s slower growth, Google’s AI talent losses, Elon Musk’s TerraFab ambitions, Revolut’s founder comp structure, and a broader question for investors and founders: which software businesses survive when AI can do more and more of the work directly in chat?
Canva, AI Cannibalization, and the Valuation Problem
Canva’s slowdown
- Canva disclosed that its 2026 growth outlook was cut by roughly a third, falling from about 30% growth to 20%.
- The company said AI feature costs are rising quickly, making it expensive to subsidize usage.
- The hosts debated whether this is just a margin issue or a sign of deeper product risk.
The bigger concern: existential risk
- The real issue is not just growth deceleration, but whether AI-native tools are eating Canva’s core use cases.
- Jason argued Canva is especially exposed because it serves the prosumer market:
- making flyers
- quick visual content
- simple web assets
- Those are exactly the kinds of tasks users now turn to ChatGPT, Claude, or other AI tools to do directly.
Private-company valuation reality
- The discussion turned to what Canva might be worth in today’s market.
- Rory suggested a rough valuation around $12B based on current growth and public-market multiples, though the real issue is whether the company is perceived as having existential AI risk.
- The hosts stressed that LPs and investors should be more skeptical of private marks, especially when growth slows and the competitive landscape changes.
Key insight
- The important question is not “Can Canva add AI features?”
- It’s: Can Canva remain the place users go to do the job at all?
- In the AI era, the danger is not just losing to a better app — it’s getting routed around entirely by the interface layer.
Google Talent Exodus and the War for AI Talent
Jeff Dean and Demis Hassabis news
- The episode covered Jeff Dean leaving Google after 27 years and Demis Hassabis stepping back from day-to-day Google DeepMind leadership.
- The hosts treated this as a major symbolic moment for Google’s AI organization.
What it says about Google
- The takeaway was not simply “Google is losing talent,” but that:
- Google’s AI efforts have been solid, but not dominant
- the company is still trying to balance:
- Cloud compute sales
- frontier model development
- moonshot science projects
- Rory described Google’s efforts as “B plus, A minus” — not A plus.
Why people leave
- The discussion emphasized that senior AI researchers often want:
- more autonomy
- more direct impact
- fewer corporate distractions
- Even at Google, some people may feel their most exciting work is being deprioritized behind the more commercial priorities of Cloud and Gemini.
Talent compensation is changing fast
- The group discussed the emergence of a “god tier” of AI compensation:
- regular employees
- AI specialists
- a tiny elite of superstars who now command huge packages
- This is especially important for late-stage companies trying to hire from Anthropic/OpenAI-level talent pools.
Key insight
- In the AI era, talent is both more expensive and more unevenly distributed.
- The best people want to work on the most interesting frontier problems — and the best companies are the ones that can give them that.
Data Centers, Regulation, and the Reality of AI Infrastructure
The political backlash
- The hosts discussed Rep. Ro Khanna proposing a “data center bill of rights” that would let local communities block AI data centers.
- The reaction on the show was skeptical: this was framed as classic NIMBYism.
Why it matters
- Data centers need:
- land
- power
- permitting
- local political support
- The group argued the real bottleneck is often power availability, not just regulation.
The likely outcome
- The consensus was that the issue will probably work itself out unevenly across states and counties:
- some places will welcome the jobs and tax base
- others will resist
- The hosts suggested communities may become more supportive if:
- electricity costs are protected
- locals receive some kind of economic dividend
Key insight
- AI infrastructure is not purely a software story.
- Compute, power, and local politics are now part of the same strategic stack.
Elon Musk’s TerraFab and the New Capital-Intensity Era
What TerraFab is
- The episode covered Elon Musk’s proposed TerraFab buildout, described as a massive, vertically integrated effort to create more domestic manufacturing and chip capacity.
- The scale of the project was framed as enormous, with a reported $16.8B initial installment and significant job creation.
Strategic logic
- The core point: Musk wants to reduce dependence on TSMC and avoid supply bottlenecks.
- Rory and Jason agreed this fits Musk’s pattern:
- huge ambition
- vertical integration
- willingness to spend aggressively when supply constraints threaten the plan
Broader implication
- The AI era is turning into a capital and supply-chain era:
- chips
- RAM
- power
- fabs
- Elon is acting like someone who believes waiting is no longer an option.
Key insight
- If supply constraints are a decade-long issue, companies that control their own capacity may win disproportionately.
- Musk is betting that control of the stack is now a strategic necessity, not a luxury.
CEO Pay, Control, and Founder Incentives
Revolut’s pay package debate
- The hosts discussed a reported incentive package for Revolut’s CEO, Nick Storonsky, that would scale with company valuation.
- The package reportedly could reach massive ownership levels if the company hits higher market-cap milestones.
Money vs control
- The conversation made a sharp distinction:
- money matters
- but often what founders really want is control
- The hosts compared this to Elon Musk’s repeated insistence that he needs control to stay engaged.
Why founder-led companies matter
- Jason argued strongly that many non-founder-run investments eventually become zeroes in this era.
- The show repeatedly returned to the idea that founder-operators are still the most important variable in high-growth tech.
Caveat: performance linkage matters
- Rory pushed for packages tied to operating outcomes, not just stock price.
- The best-designed comp packages should reward:
- real operational growth
- not just valuation inflation
Key insight
- Founder control may be a necessary tradeoff if the goal is to keep companies public and ambitious.
- But compensation should still be tied to real value creation, not just paper markups.
What AI Won’t Kill: Whatnot, Shopify, Atlassian, and the Durable Businesses
Whatnot
- The hosts loved Whatnot, the live shopping platform, calling it a clear business that people actually want.
- It was compared to:
- QVC
- Home Shopping Network
- eBay
- Their point: some businesses are not threatened by AI at all because they serve universal human behaviors like shopping and commerce.
Shopify and Atlassian
- Both companies were praised for strong quarters.
- The core lesson: if you can keep growing, the market rewards you.
- But even successful quarters may mask underlying pressure from AI and changing usage patterns.
HubSpot and SMB risk
- The hosts debated whether SMB software is under threat from:
- direct AI replacement
- increasingly strong low-end competitors
- The concern is that new AI-native startups are now appearing with real quality much faster than before.
Key insight
- Not every software company is equally exposed to AI.
- Some are being:
- replaced
- others are being repositioned
- and some are mostly benefiting from the wave
Main Takeaways
- Canva’s slowdown is a warning sign for prosumer software exposed to AI-native tools.
- Google’s AI talent losses matter symbolically, even if the business can still execute.
- The best AI talent is now extremely expensive and highly mobile.
- Data centers and power are becoming strategic infrastructure bottlenecks.
- Elon Musk is doubling down on vertical integration to avoid supply-chain dependency.
- Founder control and incentive design are becoming central governance issues.
- Some businesses, like live shopping and infrastructure software, remain resilient despite the AI wave.
- Investors should mark private companies more conservatively and stop relying on old valuation anchors.
Bottom Line
The episode argues that AI is not just changing product features — it is reordering who wins, how companies are valued, and what kinds of leadership structures can survive. The winners will likely be the companies that:
- move fast enough,
- retain top talent,
- embrace AI without being replaced by it,
- and prove real growth in public.
The losers are likely to be the companies that hesitate in 2023 and 2024 and pay for it in 2026 and 2027.