Nov 20, 2025, 10:01 AM ETConference Presentation, Fireside Chat, Interview

Peter Thiel and Softbank Sell NVIDIA - Why? & Why VC Will Hit $1TRN and The Opening of Retail

Cursor raised $2.3B at a $30B valuation (approx. $29.3M implied valuation per share context), featuring participation from Andreessen Horowitz, Thrive Capital, K2, DST, and Excel.
The "bull case" for Cursor relies on three pillars: 30–70% productivity gains for software engineers, massive Total Addressable Market (TAM) expansion, and significant multiple expansion.
TAM expansion is driven by an estimated 100–150 million active developers on GitHub alone, implying a potential trillion-dollar market at $5,000–$6,000 per developer annually.
Cursor has only 30 employees (including a PM hired 4 months ago), minimizing ESOP dilution compared to foundation model companies.
Product differentiation includes a new model that is 4–5x faster in tokens per second, allowing for higher margins and market capture.
Enterprise adoption is predicted to stabilize via Standardized ELAs (Enterprise Licensing Agreements) where Fortune 500 companies lock into a single vendor.

Key investment risks identified are profitability and durability (competition), rather than revenue growth or TAM.
Cursor faces a platform risk where its primary supplier (e.g., OpenAI, Anthropic) is also its direct competitor.
Gross margins for agentic coding tools currently lag behind traditional SaaS (70–72%); Replit and Lovable achieve >50% gross margins by using cheaper, N-1 models.
Cursor aims to reach 60%+ gross margins by eventually building and deploying its own distilled models to reduce token costs.
Moats are increasingly driven by user inertia, memory (custom linting, tool setups), and the cost of migration rather than pure technology performance.
Once enterprise users lock in, switching costs rise sharply as organizations standardize on a single platform.

Market share distribution prediction: Cursor (40–60%), Microsoft (bundling via GitHub/VS Code), and Anthropic (20%).
Cognition and xAI are considered third-tier contenders due to distinct approaches or smaller scale.
The "productivity boost" is transitioning from a luxury feature to a default necessity, with near 100% penetration expected among serious developers.
Replit V3 introduces "agents talking to agents," featuring unlimited context windows and autonomous QA capabilities, representing a step-function improvement.
Technology disruption is described as occurring in "three to four years" of rapid change before market share coalesces and stabilizes.
The "frozen skillet" analogy suggests that while performance improves 20–30x faster than historical SaaS norms, market share eventually freezes once embedded in workflows.

Oracle's credit default swaps have tripled in the last 3–4 days, signaling market concern over its leverage tied to OpenAI data center builds.
Consumer stress indicators include record-high delinquency rates on auto loans for subprime borrowers in the last 60 days.
NVIDIA faces concentration risk: two customers represent >40% of revenue, and the top four represent >50%.
A "price war" scenario where AI tools become commoditized (e.g., $100k tools dropping to $2k) is deemed "terrifying" but currently unlikely due to stickiness and integration depth.
Hyperscalers (Microsoft, Google, Meta) are "selling out" GPU capacity for two years, indicating strong near-term demand despite macro fears.
The market is viewed as being at a "red line" where a minor wobble could trigger a fast, brutal correction due to high leverage and debt.
Power availability for data centers is identified as a potential natural brake preventing total overcapacity and a demand crash.

The VC market is shifting toward a "new public market" with rapid secondary liquidity, allowing investors to trade up or down without traditional IPOs.
Late-stage VC rounds are increasingly viewed as "trading" opportunities rather than long-term ownership, with Bessemer and Kleiner leading late-stage bets.
A "bimodal" market is emerging: 4–5 massive names (OpenAI, Anthropic, xAI, Cursor, Ramp) consume a disproportionate share of capital, while the rest of the ~900 unicorns face IPO droughts.
Retail capital access is expanding via SPVs and funds of funds, potentially doubling VC asset sizes to $500B by 2030 if returns remain strong.
IPOs are becoming less attractive for top-tier companies due to high transaction costs (6–7% of raise) compared to private liquidity; public listing is reserved for when private capital becomes too expensive.
Private equity is absorbing mid-market software companies, reducing the number of public software firms to a "B-tier" of $500M–$1B revenue companies.
"King-making" by VCs is losing relevance; customer preference and product execution are becoming the primary drivers of adoption over VC backing.

Jason sent capital into GCai (AI for in-house legal), a profitable, cash-efficient company with strong customer references and low adoption barriers.
Tom Tungas predicts OpenAI will IPO in Q3 or Q4 2026, potentially pushed to mid-2027 due to alternative financing or government guarantees.
The "seed is for suckers" narrative is being challenged by a new "volume" strategy where VCs write large checks into late-stage firms to capture rapid compounding.
Early-stage exuberance persists despite public market volatility, with YC batches seeing record fundraising and demand.
A 10% "GPU guarantee" price is noted for OpenAI, suggesting potential government support for AI infrastructure.
The panel agrees that "entry price counts when TAM is unclear, but winning is the only thing that counts when time is huge."