The Twenty Minute VCElon Musk vs Sam Altman | The Implosion of Thinking Machines | Can VC Survive Public Pricing?
CHAPTERS
- 1:25 – 5:10
Public market reset: what falling multiples mean for VC returns
The conversation opens with a debate over whether depressed public-market SaaS valuations break the venture model. Rory argues multiples are behaving normally—high growth still earns premium pricing—while slow-growth names get punished. Jason frames the moment as painful for boards and portfolios anchored to 2021-era expectations.
- •Public markets are “sorting” businesses by growth rate, not declaring tech dead
- •High-growth companies can still command extreme forward-sales multiples
- •Figma as a sentiment benchmark for software valuations
- •How anchoring to peak prices distorts decision-making and morale
- 5:10 – 8:04
Is venture a ‘scam’? Converting revenue multiples into liquidity
Jason argues the core venture game is turning high revenue multiples into cash via IPO/M&A before free cash flow exists, and that an EPS world would compress fund returns. Rory reframes it as rational option-pricing on eventual category winners like Microsoft, where losers are the cost of finding trillion-dollar outcomes.
- •Venture depends on liquidity windows and multiple expansion/compression cycles
- •Backward induction: buy a basket early because you can’t wait for EPS certainty
- •Most high-multiple companies won’t become the ultimate winner—and that’s priced in
- •Why $100M ARR profitable SaaS can be ‘great’ but unfinanceable for venture math
- 8:04 – 14:15
What founders at $50–$75M ARR should do now (attach to AI or go capital-efficient)
They discuss the playbook for mid-stage SaaS companies that aren’t AI-native but still growing. Jason emphasizes urgency: build AI/agents now or risk irrelevance. Rory adds that if venture terms worsen, founders must run the business to avoid needing external capital and treat outcomes as a grind rather than a moonshot.
- •If mega-growth won’t re-accelerate, funding terms likely deteriorate
- •Operate assuming capital is no longer free; optimize for self-sufficiency
- •Find concrete AI tailwinds (agents, workflows) that map to revenue conversion
- •Even ‘grind’ outcomes can be life-changing, but don’t fit VC portfolio needs
- 14:15 – 19:48
Thinking Machines ‘implosion’: seed-stage dynamics at a $50B valuation
The team departures at Thinking Machines are interpreted as classic seed-stage founder incompatibility—just with more commas. They explore downside protection mechanics (redemption, winding down, returning capital) and why a fast partial loss can be preferable to a long turnaround attempt.
- •Founder/team breakup is the #1 seed-stage failure mode
- •Big valuations don’t change early-stage organizational fragility
- •Investor protections: redemption clauses and returning unspent capital
- •Why a quick 0.8x in 12 months can be better than a slow outcome for capital recycling
- 19:48 – 24:32
Talent wars and mission fit: why elite AI researchers don’t behave like normal employees
Jason and Rory dig into why compensation isn’t enough to retain top AI talent: researchers optimize for problems, mission, and the boss they want to work with. They connect this to portability (leaving money behind), OpenAI’s evolving comp/vesting practices, and why most startups can’t attract S-tier lab talent without a credible technical leader.
- •Top AI researchers choose intellectual problems over marginal comp
- •Portability is extreme; firms must offer the work researchers want
- •Mission-based attraction can outweigh incremental model improvements
- •Non-lab startups often need a true AI research leader as CTO/CAIO to compete
- 24:32 – 31:57
Elon Musk vs OpenAI: how a nonprofit became a for-profit—and what the lawsuit seeks
Rory lays out the origin story: charitable intent, nonprofit structure, then the realization that compute costs required a for-profit evolution. The dispute centers on whether there was fraudulent intent from day one. Rory explains Elon’s aggressive damages theory: not refunding $30M, but claiming a stake equivalent to what that seed money would buy today.
- •OpenAI’s shift: nonprofit roots → necessity of for-profit structure
- •Competing narratives: ‘donation to a charity’ vs ‘fraudulent bait-and-switch’
- •Elon’s claimed damages translate into massive dilution for current holders
- •Why the judge allowing the case to proceed creates ongoing financing overhang
- 31:57 – 44:39
Who ‘wins’ and what it does to OpenAI: discovery, depositions, and distraction risk
They argue Elon benefits regardless of outcome via discovery and public embarrassment, while OpenAI carries asymmetric downside. Jason predicts it won’t settle easily because Elon wants court and exposure; Rory believes proving fraud is a high bar, especially with juror unpredictability. They discuss how a strong GC must compartmentalize litigation so the company can keep executing.
- •Elon’s asymmetric upside: pressure, slowdown, reputational damage to OpenAI
- •Trial dynamics: unpredictable jury, bad facts on both sides, long timeline
- •Settlement is hard if one side wants spectacle more than money
- •Operational imperative: prevent litigation from consuming leadership attention
- 44:39 – 48:50
OpenAI ads: ‘inevitable’ monetization or product risk at the wrong time?
Harry presses whether ads harm the product amid competition from Gemini and Anthropic. Rory argues ads are inevitable because free-tier costs are high and consumer conversion to paid is too low; it’s a “rip the bandage off” moment. Jason counters that ads can add value when paired with high-intent discovery and kept low-density.
- •Why ads often arrive after a period of internal resistance (Google/Facebook history)
- •Free-tier serving costs + low conversion make ads hard to avoid
- •If executed cleanly, ads can be informative and improve unit economics
- •The strategic risk: introducing ads while fighting on model quality and distribution
- 48:50 – 55:03
LLMs as the new discovery layer: how big can the ad business get?
They explore the thesis that LLMs outperform Google for purchase discovery even if Google remains strong for traditional search. Jason sketches back-of-the-envelope math for massive revenue potential with relatively few ads per prompt at high CPMs. Rory connects this to intent-driven ad markets (Google/Amazon) and why OpenAI’s interface is prime real estate for complex buying decisions.
- •Discovery vs search: users shifting purchasing research into LLMs
- •Intent monetization parallels: Google search ads, Amazon retail ads
- •Ad-density can stay low while still producing significant revenue
- •Why Google’s core cash cow faces real substitution pressure in discovery
- 55:03 – 1:01:39
AEO/GEO (Answer Engine Optimization): who captures value and Adobe’s Semrush move
The discussion turns to second-order winners: tools that help brands appear in LLM answers (AEO/GEO) versus tools that buy paid ads. Rory argues OpenAI will likely keep most paid-ad value directly, but AEO is a real emerging category. They interpret Adobe’s Semrush acquisition as a strategic bet to own the enterprise distribution path for AEO-style products.
- •Two layers: optimize for ‘free answer’ visibility vs manage paid ad buying
- •Skepticism about early GEO tooling quality vs belief the category will be big
- •Adobe buying Semrush as a distribution-heavy, enterprise-friendly AEO wedge
- •Why big incumbents prefer buying scale/revenue over tiny frontier startups
- 1:01:39 – 1:10:12
Mega-rounds analyzed: ClickHouse at $15B and underwriting growth persistence
They analyze ClickHouse’s rise: an older open-source OLAP product that caught AI tailwinds, nailed monetization, and became infrastructure for AI workloads. Rory explains late-stage pricing as a bet on growth persistence and category size, with valuation risk expanding once technical and go-to-market risks are reduced. They compare the OLAP category’s potential to Snowflake/Databricks outcomes.
- •ClickHouse: open source origins → hosted product → AI-driven demand spike
- •Late-stage logic: pay up when you believe growth persists for 2–3+ years
- •Category framing: OLAP/analytics databases vs transactional/data warehouse markets
- •Downside case: market smaller than expected or growth decays faster than priced
- 1:10:12 – 1:11:58
Competitive investing fades at the ultra-late stage: Sequoia in both Anthropic and OpenAI
Harry asks if competitive investing norms are collapsing as top firms back multiple leaders. Rory argues at $350B-pre type rounds, investors often have minimal rights and little strategic access, making conflicts less meaningful—more like public-market positioning in private assets. Jason agrees the playbook is to own a lot early or buy exposure to ‘guaranteed winners’ late if LP capital is available.
- •Information rights and board access define conflict severity—often absent ultra-late
- •Multi-stage firms use late rounds to ‘clean up’ early-stage misses
- •LP appetite and SPVs enable massive follow-on exposure to category leaders
- •Venture increasingly resembles public growth investing—just with private pricing
- 1:11:58 – 1:23:12
Replit at $9B: product step-change, agent workflows, and valuation justification
They close by debating Replit’s valuation, with Jason emphasizing that the product is dramatically better than at prior rounds—moving from “80% finished” projects to workable agent-driven outcomes. Rory agrees it’s powerful but notes the bet still hinges on growth persistence. They reflect on how the interface abstracts code away, signaling a shift from programming to specification.
- •Valuation step-up tied to both ARR growth and qualitative product transformation
- •Agents unlock completion and reliability, not just incremental UX improvements
- •Shift from coding to describing: abstraction hides code to boost throughput
- •Core underwriting question remains: can current growth rates persist long enough?
