20VC: Will SpaceX IPO at $1.5TRN | Will Cursor Kill Figma | Lightspeed Raises $9BN | OpenAI: $1BN from Disney, New CRO & #1 App in App Store | Oracle and Broadcom Hit: Now the Time to Buy?

Analysis of Lightspeed's massive $9B raise across six funds ($2B for early/venture, $7B for growth). Discussion centers on how mega-funds destroy seed pricing discipline and create insurmountable advantages through portfolio construction - combining early wins (Rubrik, Navan) with late-stage concentration bets (Anthropic). The hosts debate whether this makes small seed funds irrelevant.

• •Lightspeed's success formula: early stage companies that mature over 10-12 years + stuffing billions into hot late-stage deals like Anthropic
• •At $9B scale, seed pricing becomes irrelevant - they can pay 30-40-50M pre-seed because it's just an entry ticket to later rounds
• •LP appetite for pure seed managers ($50-100M funds) has faded significantly since 2021
• •The real game is getting $200M+ to work at 6-7x returns, not optimizing seed entry prices
• •Private markets keeping winners like SpaceX/Databricks/OpenAI away from public investors is 'the greatest gift of venture in our lifetimes'

Disney's $1B investment in OpenAI is actually a cross-licensing deal allowing OpenAI to use Disney IP for image generation. This represents the first major template for how content owners will monetize their IP in the AI era, with Disney simultaneously sending cease-and-desist to Google for unlicensed use. The deal signals the end of AI companies freely scraping content.

• •The $1B is 'round-trippy' - Disney gets equity, OpenAI gets licensed access to Disney characters for image generation
• •This creates a pricing template for other content owners - first movers get better deals, rates will ratchet up over time
• •ChatGPT became the #1 most downloaded app in the US App Store for 2025, reaching 800M users
• •User growth has slowed to single digits (running out of humans), raising questions about whether OpenAI follows the Robinhood model (cross-sell existing users) or Meta model (reach 4B+ users)
• •The real question: can OpenAI maintain growth rates as it approaches saturation, or does it get valued on cash flow instead of growth?

Oracle shares plunged 45% from September highs after spending $12B in quarterly CapEx (vs $8.2B expected) to build data centers for OpenAI. The market initially loved the revenue commitments, then realized the capital intensity destroys margins. Oracle and CoreWeave represent the 'high octane' AI infrastructure plays - massive upside if AI spending continues, but extreme downside as marginal commodity providers.

• •Oracle's pop on RPO announcements was a 'sugar high' - the market didn't price in the CapEx required to deliver that revenue
• •Oracle and CoreWeave are 'weak guys' with no IP, high risk of margin compression, making them volatile AI bellwethers
• •CoreWeave down 60% from July highs, Oracle down 45% - these are the canaries in the coal mine for AI infrastructure spending
• •They represent amplified bets: if AI CapEx continues strong for 2+ years, they could double; if it plateaus, significant downside
• •The market is 'interrogating for truth' - filtering which companies have real AI plans (Google, Microsoft, OpenAI) vs those without clear ROI

Broadcom lost $300B in market cap over concerns that a $21B chip order from Anthropic will compress margins. The fundamental issue: Anthropic is buying custom chips from Broadcom specifically to avoid paying NVIDIA's 75% gross margins. This is a 'made-to-order' business with lower defensibility than NVIDIA's branded architecture, even though Broadcom still trades at high-teens revenue multiples at $1.6T market cap.

• •Anthropic's entire strategy is avoiding NVIDIA's premium pricing - they won't pay Broadcom full retail either
• •Broadcom is doing 'made-to-order' custom chip design vs NVIDIA imposing their architecture on customers - fundamentally different defensibility
• •Despite the drop, Broadcom still worth $1.6T at high-teens sales multiple - 'not like everything went cheap'
• •The market gives different companies different 'gross margin passes' - trying to figure out who deserves it and who doesn't
• •Broadcom's story mirrors Oracle: initial excitement about revenue, then reality of margin compression sets in

Cursor launching AI design tools represents massive category convergence - design, prototyping, and production coding are collapsing into single platforms. The insight: everyone wants to talk to the same AI agent (designers, product, engineers, DevOps) rather than being fragmented across 11 different tools. This threatens Figma's position and represents a broader trend of AI destroying traditional software category boundaries.

• •Marketing, sales, and support have already converged in e-commerce (Klaviyo bringing in Workday co-CEO to adapt to this reality)
• •The disconnect between design and code output is 'the most jarring thing' - you can spot 'vibe coded' sites from YC batch by their Claude artifacts look
• •Cursor reached $500M revenue in 9 months vs Lovable/Replit's combined ~$500M in a year - professional software budgets are 10x larger than prosumer
• •55% of all enterprise AI spend is on coding and coding-related tools (per Mendelow data) - this is the epicenter of enterprise AI
• •Winner will be determined by 'who wants it most' and works 12/12/9 - everyone can copy features in weeks now, not quarters

The real risk for incumbents isn't being completely displaced - it's being 'maimed' where existing customers don't churn but new customer acquisition slows and seat expansion stops. Companies like Figma, GitLab, and Atlassian face this: logo retention stays good, but NRR drifts down as teams use Cursor alongside them, and new YC companies defer purchases. This slow bleed is more insidious than outright disruption.

• •Maiming looks like: 15,000 customers to 15,200 (minimal growth), existing retention good but NRR declining, new generation defers purchase
• •GitLab 'arguably being maimed', Atlassian slowing down, even MongoDB should be growing 50% given AI explosion but has too many competitors (Supabase, etc)
• •UiPath is the poster child: not hemorrhaging customers, back to 16% growth, stock up 27%, but got maimed by RPA being replaced with agents
• •The fix requires 'supreme act of CEO will' - refounding the company in the age of AI, not just intellectual understanding but driving it through organization
• •Companies with 98% GRR and 107% NRR (like UiPath) have 18-24 months to build agentic products their base will buy before time runs out

Boom Supersonic, the YC-backed supersonic plane company, is raising $300M to sell jet engines to data centers for power generation (via Crusoe order). The company went from $1B valuation in 2021 to a ~$500M crash, now back to $1.5B by pivoting to AI infrastructure. This mirrors how Rolls Royce and GE already sell airplane engines for power generation - it's actually easier than building planes.

• •Boom's journey: $1B valuation (2021, no revenue) → $500M crash (Dec 2024) → $1.5B (2025) by adding data center power business
• •All airline engine manufacturers (GE, Rolls Royce) already do this - the bulk of airplane engine tech translates to generators
• •It's 'a lot easier to plop an engine on the ground for electricity than put it in a plane for propulsion'
• •This represents the ultimate 'co-attaching to AI budget' - pivoting a hard-tech company to ride the infrastructure wave
• •The whiplash for investors: massive valuation swings based on which narrative (planes vs power) the market believes

SpaceX is rumored to IPO at $1.5T valuation despite doing only $15-16B in revenue (78x 2026 revenues). The hosts introduce the 'Elon Option Value' (EOV) concept: you can't justify the numbers on fundamentals, but Elon has pulled trillion-dollar rabbits out of hats three times (Tesla, SpaceX/Starlink, OpenAI). The premium is betting he'll find another trillion-dollar market not in the original plan. Space-based data centers mentioned as potential next play.

• •You cannot run the numbers to get $1.5T - it requires applying 'Elon Option Value' on top of a 10x revenue multiple
• •Elon's track record: PayPal, Tesla ($1T+), SpaceX rockets, SpaceX/Starlink communications, OpenAI - most impressive entrepreneurial record in 30-40 years
• •Tesla IPO'd at $1.7B in 2010; SpaceX will IPO at $1.5T - that's the difference in staying private longer (1000x more value captured by VCs)
• •Founders Fund owns 10% from backing Elon after Peter Thiel fired him from PayPal but 'treated him like a human being' - could be worth $150B
• •The IPO challenge: raising $30B (2% dilution) vs typical 8% ($120B) - is there $120B of risk capital willing to buy 70x revenue space investments?
• •Space-based data centers could be the next big play - 'he's the only one that can build data centers in space' and it's easier than Mars return trips