newsfilter.io
Interview, Fireside Chat

David Weisburd, 10X Capital: Finding Alpha and LP Bias

  • The "Liquidity" segment featuring Jason Calacanis is expected to reach over 10 episodes by the time of the podcast release.
  • AI is predicted to evolve from a distinct investment category into a catch-all term for innovation, rendering separate "tech" sector classifications obsolete.
  • The venture capital landscape for AI is projected to progress through an infrastructure phase, followed by verticalization (e.g., legal or healthcare AI), and finally reach a stage where AI models are non-commoditized.
  • Strategic moves by leaders like Sam Altman to create an "app store for AI" are anticipated to be key drivers in preventing AI model commoditization.
  • Power law returns are asserted to end at approximately $100 million valuations, making 1000x returns mathematically impossible for single positions at this level.
  • Disruption by Open AI is considered acceptable only if the 10% probability of success results in a trillion-dollar expected value; conversely, a 90% disruption risk is unacceptable regardless of the outcome.
  • Founders are advised to evaluate their exposure to Open AI disruption probabilistically, distinguishing between unacceptable 90% risks and acceptable 30% risks for potential $200 million outcomes.
  • Large entities like Open AI are not expected to target companies valued at $10 billion or $100 billion, though the "paperclip issue" poses a risk of accidental business termination.
  • As funds scale, they are projected to attract major institutional capital, including sovereign wealth funds and CalPERS, particularly when fund sizes approach $5 billion, enabling checks of $500 million.
  • A new $75 million solo GP fund is forecasted to generate DPI (Distributions to Paid-In Capital) specifically in the years 2033 or 2034.
  • The period between 2024 and 2033 involves potential risks including management complacency, addiction to fees, partnership dissolution, and forced founder returns to office environments.
  • Historical top-quartile performance in one vintage does not guarantee similar results for the 2034 vintage.
  • GPs are advised to sell 10% to 20% of positions with 50x to 100x returns to provide capital to LPs, balancing those seeking 12% predictable returns against those pursuing high risk.
  • Early-stage VCs are recommended to maintain a portfolio of 15 companies to balance boldness, whereas 10 positions are less aggressive and 25 positions increase the odds of a 3x fund while reducing the odds of a 20x fund.
  • A competent solo GP managing a $10 million fund with 25 companies is projected to have high odds of a 3x outcome but low odds of achieving a 20x result.
  • Investors generally do not underwrite an 8% return unless specifically targeting top-quartile SoftBank portfolios.
  • The industry is predicted to see increased proactive secondary strategies and asset sales to manage DPI, necessitating humility from GPs to counter LP short-term greed.
  • It is deemed highly rational to sell portions of a position appreciating 100x to assist in raising the next fund rather than retaining all assets for potential carry.
  • In scenarios where a position has limited upside potential (e.g., $1.5 billion value against a $1 billion valuation) with high risk, the recommendation is to sell the entire position.
  • Founders in turbulent markets are encouraged to assess disruption probabilistically rather than fearing giants like Google or Open AI without specific data.
  • DPI is characterized as a "great equalizer" and "ground truth" but remains a highly lagging indicator that fails to capture fund trajectory in the years immediately following an investment.
  • LPs exhibit a systematic bias toward existing managers, re-upping automatically unless fraud occurs or returns are completely absent.
  • The "myth of jurisdiction" is expected to persist as managers use diversification to increase fees, despite academic research indicating that diversification beyond 15 positions is inefficient for generating alpha.