The Twenty Minute VCOpenAI's Multi-Billion Deal with AMD & Polymarket, Vercel and Supabase Raise Mega Rounds
CHAPTERS
- 0:00 – 5:42
OpenAI–AMD chip deal: warrants, leverage, and who has power in the stack
The group unpacks OpenAI’s supply partnership with AMD, focusing on the headline-grabbing warrants that could amount to up to ~10% of AMD. They argue the structure reveals a clear power hierarchy: OpenAI can demand upside from weaker suppliers while conceding equity upside to stronger ones like NVIDIA.
- •OpenAI’s AMD agreement includes warrants tied to buying/shipping chips and AMD’s stock performance
- •Contrast with NVIDIA: strong vendors can extract equity from OpenAI; weaker vendors may have to give equity to OpenAI
- •Warrants as a “kingmaker package” to avoid reputational risk and ensure supplier commitment
- •Market reaction (AMD stock pop) illustrates how OpenAI can “bestow” market cap despite heavy cash burn
- 5:42 – 9:46
The Wintel analogy: OpenAI as Microsoft, NVIDIA as Intel, AMD as the second source
Rory frames the situation as history rhyming with the DOS/Windows–Intel era. OpenAI is positioned as the user-facing monopoly-in-formation, NVIDIA as the critical compute counterpart, and AMD reappears as the strategically necessary second supplier—while Microsoft plays an IBM-like role in enabling a future rival.
- •Historical parallel: Microsoft/Intel dominance and AMD as second-source supplier
- •OpenAI + NVIDIA as a modern duopoly controlling the key layers (users + chips)
- •Strategic importance of a second source even when one supplier dominates
- •Microsoft’s risk: enabling a “monster” that competes with its platform ambitions
- 9:46 – 16:59
How NVIDIA reacts: margin inversion, component economics, and chip monopoly dynamics
The conversation shifts to how NVIDIA likely views OpenAI’s diversification and why Jensen can tolerate some share loss. They discuss the unusual inversion where the component vendor captures extraordinary margins, why buyers want to attack that margin, and what could change if GPUs ever became a more commoditized market like memory.
- •NVIDIA’s profits and 50%+ margins create inevitable pressure for competition
- •Component vendors usually get squeezed (“cost plus”), but architectural lock-in flips leverage
- •OpenAI’s leverage comes from users/utilization demand, even while burning cash
- •Risk scenario: if GPU supply becomes commodity-like (memory market dynamics), margins collapse
- 16:59 – 23:58
DevDay apps in ChatGPT: why the demos felt like “Slack integrations” all over again
They evaluate OpenAI’s DevDay push to run third-party apps inside ChatGPT (e.g., Canva, Zillow). Jason is underwhelmed, arguing integrations often exist but rarely become habitual workflows—much like Slack’s app ecosystem—unless there’s a truly jaw-dropping use case.
- •Desire: talk to business apps directly (e.g., CRM) without logging into each tool
- •Skepticism: integrations may be ‘nice’ but not a superior UI to purpose-built apps
- •Analogy: Slack had connectors everywhere, yet few users actually ‘used apps inside Slack’
- •OpenAI may win mindshare, but “death of apps” is unlikely
- 23:58 – 25:11
AgentKit and ‘easy agents’: does OpenAI commoditize agent platforms or not?
The group distinguishes between lightweight agent demos and the complexity of enterprise-grade orchestration. They suggest OpenAI could wipe out simpler automation/agent tooling, but many real deployments will still require significant product surface area, management layers, and ongoing customization.
- •AgentKit demo speed vs. real enterprise complexity
- •Enterprise agents likely require orchestration, governance, monitoring, and integrations
- •OpenAI may provide primitives/connectors and leave deeper tooling to specialists
- •Unclear how much OpenAI will prioritize agent ops vs. “bigger fish”
- 25:11 – 33:28
Mega-round venture math: a $1B raise at $5B pre and the ‘forgiving’ nature of VC pricing
They debate whether huge early checks at lofty valuations break the venture model. Rory argues infrastructure/deep-tech markets reward proven operators who can credibly solve hard problems, while also noting VC is more forgiving than PE/public markets—though not infinitely forgiving.
- •Why certain founders can raise massive rounds: scarcity of proven deep-tech builders
- •Compressed diligence logic: assume founder picks the right problem and can execute
- •VC vs PE: VC is most forgiving on entry price due to power-law outcomes
- •Risk: if outcomes cap at ~5x while only a fraction become ‘wildly amazing,’ fund returns suffer
- 33:28 – 37:18
Why comps can mislead: valuation relativism vs. investing for 7-year outcomes
Rory pushes back on using comparable valuations to justify prices, arguing comps answer what something is worth today—not what it will be worth in seven years. They connect this to 2021-style overpayment and the danger of assuming high revenue multiples will persist over a decade.
- •Comps are a banker tool for ranking today, not forecasting future value creation
- •2021 lesson: “cheap relative to crazier comps” can still be a bad investment
- •To underwrite big prices, you must believe in durable multiple regimes over time
- •Founder confidence games: seeing $100B outcomes up close shifts perception of what’s ‘plausible’
- 37:18 – 42:34
LP liquidity and secondaries: universities selling stakes and why liquidity can vanish
The group discusses endowments and LPs selling venture positions, and Jason shares a first-hand story about how restrictive LP transfer rights can be. They argue more secondary liquidity would benefit the ecosystem, while warning that liquidity disappears due to fear—not lack of money—reinforcing why public markets still matter.
- •University endowments face added pressure to hold more liquid portfolios
- •LPs often have little/no right to sell without GP consent; secondary processes can be frictionful
- •Longer private timelines increase the need for structured liquidity (LPs and employees)
- •Liquidity risk: it evaporates when buyers get scared and hoard cash
- 42:34 – 53:04
IPO bar in 2025 and the Snyk case: scale thresholds, PE interest, and ‘take control of destiny’
They assess what it takes to IPO now, citing surprisingly high median revenue run-rates among recent IPOs, and debate whether Snyk’s growth makes it borderline. The conversation turns to the scarcity of PE bids for ‘good-but-subscale’ assets and what boards must do if liquidity options are weak.
- •IPO market realities: higher scale expectations; smaller issuers risk being ignored
- •PE valuations and comps suggest materially lower outcomes vs prior private marks
- •Jason’s concern: PE phones aren’t ringing like they used to—even for healthy Rule-of-40 assets
- •Board playbook: re-incent founders (equity-for-growth), drive profitability, and build a second act (often AI-linked)
- 53:04 – 58:02
Founder longevity and CEO transitions: when swapping CEOs works (and when it doesn’t)
They discuss the rarity of founders who run 15-year journeys and the mixed record of replacing them. Rory argues CEO swaps can help if product-market fit is already achieved and the issue is operational management—but almost never works as a path to finding PMF.
- •Founder endurance is uncommon; most people aren’t wired for decade-plus marathons
- •CEO replacement can work post-PMF if the founder is weak operationally
- •Replacing a founder to ‘find PMF’ is usually self-delusion; professional managers rarely do that
- •The “would you sell for $500M?” question is dismissed as low-signal and often misleading
- 58:02 – 1:04:59
Vercel & Supabase mega-rounds: ‘Captain Obvious’ dev infra bets and the ‘suicide round’ debate
They evaluate Vercel’s $300M round at ~ $9B and tie it to Supabase’s momentum, arguing both sit on a clear developer migration and the explosion of app creation. They challenge the ‘suicide round’ framing, noting rapid growth can justify stable revenue multiples—while acknowledging the risk of hitting market-size walls.
- •Vercel + Supabase as structural infrastructure primitives for the new app-building wave
- •‘Captain Obvious’ strategy: big obvious trends beat overcomplicated theses
- •Why multiples can stay constant across rounds if growth persists unusually long
- •Primary risk: abrupt deceleration or TAM ceiling causing valuation compression; plus needing more cash later
- 1:04:59 – 1:18:58
‘Kingmaking’ in VC: how capital, prestige, and token costs change competitive dynamics
A debate breaks out on whether venture investors ‘kingmake’ winners, especially with rapid successive rounds early in a company’s life. They land on a nuanced view: early execution matters, but capital and brand can harden advantages—particularly in AI markets where compute/token costs and balance-sheet credibility can be decisive.
- •Prestige + capital can deter entrants and accelerate follow-on rounds (“momentum flywheel”)
- •Counterexample: Legora vs Harvey shows well-funded leaders can still be challenged with better product
- •AI markets may invert classic SaaS: capital becomes more useful later due to token/compute burn
- •Founder advice: if offered attractive ‘strategic’ capital in competitive markets, game theory often says take it
- 1:18:58 – 1:26:34
SPACs returning and Polymarket’s $2B investment: deregulation, incentives, and strategic money
Rory argues Chamath’s new SPAC terms are closer to rational by tying sponsor economics to performance rather than deal completion. Then they tackle the Polymarket headline—framing it as a sign of shifting regulatory winds and strategic capital, with ICE/NYSE investing for market structure and distribution advantages rather than pure VC returns.
- •SPAC incentive reset: sponsors earn only after meaningful stock performance (less misalignment than 2021)
- •Regulatory context matters: what was constrained/illegal can become investable quickly under new regimes
- •Polymarket: strategic investment likely implies exclusivity, data, distribution, and platform integration
- •Prediction markets: question is whether non-sports use cases become a durable, legitimate category
- 1:26:34 – 1:32:36
Quick-fire: Tim Cook odds and vibe-coding ARR forecasts for Replit & Lovable
They close with rapid takes: Tim Cook leaving Apple this year is seen as unlikely absent health issues, though succession planning is expected. On Replit and Lovable, Jason leans toward hitting ambitious ARR targets, emphasizing how quickly products improve even if ‘looky-loo’ demand flattens.
- •Apple succession planning is normal; near-term CEO exit is improbable without external shocks
- •Vibe-coding platforms may see demand flatten as casual users churn, leaving stickier prosumers
- •ARR outlook depends heavily on fast product iteration—not traditional SaaS maturation curves
- •The rate of improvement is a key variable when forecasting in AI-native categories
