The Twenty Minute VCCanva Slahes Growth | Talent Exodus at Google | Revolut's $50B CEO Package | Musk's $55B Terrafab
CHAPTERS
- 0:00 – 6:32
Canva’s growth slowdown and the real issue: AI cost + product relevance
The group opens on Canva cutting its 2026 growth outlook from ~30% to ~20%, citing exploding AI serving costs. They quickly move past the headline to the deeper question: is AI just a feature Canva can absorb, or a platform shift that makes prosumer creative tools structurally less necessary?
- •Canva revenue scale and decelerating growth context (still large, still growing)
- •AI serving costs and margin pressure as a near-term explanation
- •The bigger risk: are they on a path from 30→20→10 because workflows move into ChatGPT/agents?
- •Prosumer tools are more exposed than enterprise workflow products
- 6:32 – 8:26
Feature vs platform shift: why agents may route around Canva entirely
Jason argues the scariest signal isn’t churn—it’s that agents don’t even propose tools like Canva/Notion anymore. The conversation expands into “Fortnite-ification”/bundling: if ChatGPT becomes the universal interface, standalone tools lose UI real estate even if they remain capable products.
- •Agents and assistants as the new default ‘chooser’ of tools and workflows
- •Bundling kills standalone subscriptions when value is ‘good enough’ inside ChatGPT
- •Prosumer behavior shifts faster because everyone is ChatGPT-fluent
- •UI disaggregation risk: users ask for outcomes, not apps (Uber analogy)
- 8:26 – 10:22
Canva’s strategic response: build in-house models—too late?
Rory lays out the bull case: Canva can reduce AI costs by shifting from frontier providers to in-house models, potentially restoring unit economics. But even if costs drop 80–90%, the central question remains whether Canva can rebuild habit and distribution in a world where users start inside AI-native interfaces.
- •In-house model development as a path to cheaper serving costs
- •Debate: why didn’t Canva do this earlier? Timing risk is compounding
- •Even with cost parity, the distribution/UI battle may be lost to AI assistants
- •Product challenge: match ChatGPT-level accessibility for the prosumer segment
- 10:22 – 18:00
Public vs private: who really benefits from delaying an IPO?
They debate whether staying private is a blessing, concluding it depends on whose incentives you’re optimizing for (founders/employees vs VCs seeking liquidity). The group notes that public markets amplify pain, but the platform-shift problem exists either way; narrative control and proving growth become paramount.
- •IPO timing as a liquidity question for VCs more than founders
- •Private companies can ‘hide’ longer, but liquidity dries up when appetite dies
- •Founders may prefer long-term mission over opportunistic exits
- •Ultimately: markets follow fundamentals—growth is the only proof of non-obsolescence
- 18:00 – 21:49
Who caught the AI wave vs who benefited from being well-positioned?
Harry asks for examples of companies that moved fast into AI; the discussion distinguishes “being positioned” (infra winners) from true reinvention. They cite Intercom and Palantir as standout cases, while noting how hard it is for large incumbents to pivot compared to smaller, faster-moving teams.
- •Intercom as a reinvention example (hard journey)
- •Infra winners (Datadog/Cloudflare/etc.) benefited from demand surge more than reinvention
- •Coding tools as the most visible category of rapid iteration
- •Palantir as an app-layer example with dramatic re-acceleration
- 21:49 – 25:09
Palantir’s playbook: FDEs + outcome-based deals as the wedge
Palantir’s surge is attributed to deep deployment capability (field engineers) and willingness to sell outcomes, not just software seats. They argue most enterprises can’t execute that shift because it requires rare talent, credibility, and organizational muscle.
- •Field deployment expertise as a durable advantage in enterprise AI adoption
- •Outcome-based pricing as a real (rare) commitment vs a marketing slogan
- •Enterprise AI needs operators who can drive change, not just models
- •Why most legacy SE orgs can’t match Palantir’s AI execution velocity
- 25:09 – 33:55
Valuation reality check for LPs: growth rates, existential risk, and secondaries
The conversation turns to LP guidance: you must anchor to growth and durability, then adjust for existential risk perception. They discuss mark discipline, the value of trimming via secondaries in boom times, and the power-law tradeoff (selling early vs compounding rare outliers).
- •Anchor valuation to actual growth + forward durability, not prior marks
- •Existential risk changes multiples even at similar growth rates
- •Secondaries: emotionally hard but can lock wins when windows are open
- •Power laws: trimming hurts most on the few biggest winners
- 33:55 – 44:01
Google’s talent exodus: Jeff Dean leaves, Demis steps back—what it signals
They analyze departures of top AI leaders as both a morale and priority signal: Google’s compute allocation incentives favor cloud monetization and competitive frontier models, leaving ‘science moonshots’ deprioritized. For researchers, leaving for well-funded independent labs may be more appealing than internal politics.
- •Stock reaction underscores how individual-centric AI leadership still is
- •Internal priority conflicts: Cloud margins vs Gemini competitiveness vs science bets
- •Rationale for leaving: autonomy + funding for ‘AI for science’ labs
- •Google’s performance graded as solid but not ‘A+’, especially in coding
- 44:01 – 46:40
Why Anthropic retained talent: mission alignment + perfect financial timing
They contrast Google’s constraints with Anthropic’s ability to offer both mission and exceptional economics. The panel argues Anthropic is positioned to IPO soon, given peak narrative strength and the signaling effect such an IPO would have on the entire talent and capital market.
- •Mission framing + PBC structure as a retention tool
- •Extraordinary equity outcomes distort the whole hiring market
- •IPO timing as ‘brass ring’ moment while momentum is strongest
- •Anthropic/OpenAI comp becomes the benchmark competitors must address
- 46:40 – 50:47
The ‘three-band’ comp market: regular, AI, and god-tier superstars
Jason describes a new compensation stratification: baseline employees, AI specialists, and a small ‘god tier’ of 1–5 superstars requiring seven-figure packages and outsized equity. This creates cultural strain but is framed as necessary to build skunkworks capable of real AI product leaps.
- •Three compensation bands emerging across tech organizations
- •Skunkworks + ‘god tier’ hires as prerequisite for true AI reinvention
- •Team morale and fairness tensions as comp dispersion widens
- •Anthropic/OpenAI secondaries and valuations intensify the arms race
- 50:47 – 56:13
Data center backlash and NIMBY risk: politics, incentives, and power constraints
They discuss growing political resistance to data centers (including proposals for local veto power) and whether it slows deployment. The group argues opposition is partly about opacity and local costs (electricity), and could be mitigated via clearer community benefits—though power availability is the tighter constraint.
- •Local backlash framed as a competitiveness risk vs China
- •Resistance often driven by unclear benefits and perceived local downsides
- •Potential mitigations: guarantee no rate hikes, offer local dividends/benefits
- •Regulatory competition across US states may keep buildout moving
- 56:13 – 59:06
Musk’s $55B ‘Terrafab’ bet: vertical integration to escape TSMC constraints
They assess Musk’s fab ambition as a response to long-dated supply chain bottlenecks in chips and compute. The upside is strategic independence; the downside is that an ‘all-in’ project is most vulnerable if AI capex demand slows.
- •Strategic driver: capacity scarcity and dependence on TSMC allocation
- •Extreme vertical integration across energy, robotics, and fabrication
- •Macro risk: any AI spend slowdown hits the biggest bets first
- •Intel’s capital moves as a broader signal of the AI capex cycle
- 59:06 – 1:16:46
Revolut’s $50B CEO package: incentives vs control and shareholder dilution
They dissect Revolut’s reported ratcheting CEO package that could lead to ~40% ownership, debating whether it’s about money or control. Rory argues such packages must be tied to operational metrics, not just stock price, and highlights the shareholder protection challenge when dilution becomes open-ended.
- •Ratcheting equity packages: rare but increasingly visible for founder-CEOs
- •Control vs cash: dual-class voting isn’t always enough to prevent ouster pressure
- •Participation rate math: how much value creation goes to the CEO
- •Need for operational performance gates, not only valuation targets
- 1:16:46 – 1:20:33
Whatnot’s $20B raise: a reminder of non-AI categories that still scale
The panel welcomes a break from AI and praises live shopping as a durable human-behavior business model (QVC for the internet). They argue investors should study sectors not directly ‘eaten’ by ChatGPT, where growth can still be driven by commerce, finance, and other fundamental needs.
- •Whatnot’s model: live commerce + take rate on growing GMV
- •Why live selling works (historical precedent: QVC/HSN, eBay dynamics)
- •Opportunity thesis: look for categories less exposed to AI UI bundling
- •Valuation humor: AI multiples vs commerce fundamentals
- 1:20:33 – 1:29:42
Atlassian/Shopify results, Loom monetization, and the ‘stress signals’ in SaaS
They review strong quarters from Atlassian and Shopify, emphasizing that markets reward demonstrated growth when existential narratives loom. But Jason flags monetization moves (cutting free Loom seats) as a sign of operational pressure, and they close on competitive threats to SMB incumbents like HubSpot from AI-native entrants.
- •Markets follow proof: growth is the antidote to ‘dying’ narratives
- •Monetization/harvesting (e.g., Loom free seats) as a stress indicator
- •Seat-based SaaS faces ‘permanent assault’ from agentic workflows
- •SMB incumbents threatened less by DIY coding and more by fast, AI-native new entrants
