Nov 13, 2025, 10:00 AM ET

Michael Burry Shorts NVIDIA and Palantir & Has Defensibility Died in a World of AI?

The venture ecosystem is shifting from "Co-pilot" tools to AI agents that function as integrated team members with infinite context and autonomy.
Early-stage defensibility is eroding; founders can now launch 10+ competitive clones within 30 days due to rapid AI-enabled development.
Jason Fried observes that while the pace of evolution is "brutally competitive," revenue is accelerating significantly, with OpenAI targeting $20B ARR and Anthropic projecting $70B by 2028.
Michael Burry's $1.1B short position on NVIDIA and Palantir is analyzed as a high-risk timing bet that requires the stock to drop below specific thresholds (e.g., $160 for 2x return in 47 days) to be profitable.
Jason Fried argues that shorting AI capex is a "zero-sum" game where timing is more critical than directional correctness, making it a net-negative strategy for most investors despite the eventual likelihood of a correction.
Harry Stebbings notes that while AI CapEx may eventually overshoot, the current revenue traction is undeniable, rendering skepticism about the trend's existence "overly negative."
The investment thesis is moving toward "vertical specialization" where data moats accrue over time (e.g., patent law algorithms improving with data ingestion) rather than relying on first-mover advantage in horizontal markets.

Sequoia Capital executed a leadership transition where Ron Cohn stepped down as Global CEO after a three-year tenure, replaced by Pat Grady (growth) and Alfred Lin (early stage).
This leadership split is interpreted as a strategic move to address internal dissatisfaction with deal flow and a need to better compete in the AI sector.
The hosts agree that many executives and VCs from the last decade are no longer suited for the current market pace, suggesting a massive shakeout in the industry.
The partnership model in venture capital is identified as inherently dysfunctional, making it difficult to tie individual economic performance to economics and often leading to internal friction.
Harry Stebbings highlights Hummingbird Capital as a standout outlier, achieving a $1B+ position in a single IPO (Blink) with a $150M fund size through high capital efficiency.

The current fundraising environment is described as "binary," where only companies with top-quartile AI-native growth or elite brand backing (e.g., YC, Stripe) secure funding; mid-tier SaaS companies face extreme difficulty.
Founders are advised to avoid formal "processes" when they already have strong interest from key investors; the optimal strategy is nurturing relationships so that a term sheet can be issued immediately without a data room.
Harry Stebbings argues that founders running a formal process when a top-tier investor offers a term sheet immediately are often prioritizing price optimization over partner selection or playing a "game" that risks losing the lead.
Jason Fried notes that the "accidental process" of sharing data with investors who are not ready to commit is a common founder mistake that can jeopardize future deals.
The consensus is that successful fundraising now requires either attaching to the massive compute budget or demonstrating the ability to replace human labor; "stealing" revenue from incumbents is viewed as the hardest path.

Datadog: The stock surged 23% as the company successfully co-attached to the AI compute boom, selling infrastructure to hyperscalers rather than competing in end-user AI apps.
Clio: The legal tech company raised at a $5B valuation by pivoting to include AI and fintech payments, illustrating the importance of finding new revenue streams beyond legacy models.
Duolingo: The stock dropped 25% because the market viewed its AI integration as merely improving the product rather than attaching to compute budgets or replacing humans at scale.
Gamma: The presentation software raised $100M at a $2.1B valuation after hitting $100M ARR with 50 employees, demonstrating the revenue potential of AI-native, low-headcount companies.
Replit: The host launched 10 apps in 125 days using AI agents, noting that the Replit V3 agent has evolved into a "team member" capable of handling high-value, autonomous tasks without human engineering.

Increased market variance suggests seed funds may need to increase deal counts (e.g., from 20 to 40) or raise larger funds to maintain sufficient diversification, unless they accept lower ownership percentages in exchange for higher outcome multiples.
The hosts discuss the trade-off between "high ownership" strategies (e.g., Lightspeed) and "high multiple" strategies (e.g., Hummingbird), noting both are viable but require different capital deployment tactics.
Maintaining significant ownership in early-stage companies into a $10B+ exit is identified as an "S-tier" skill that requires extreme capital efficiency and strategic follow-on discipline.
The "one-in-three" probability of a startup succeeding at the Series B stage is debated, with some arguing that adjacent competition from giants like Salesforce and Canva increases the risk to "one-in-seven."
The hosts conclude that the most "exciting" time in software since the 1990s is happening now, driven by fundamental changes in how work is done, rather than just incremental UI improvements.