Bloomberg Television13 min
Anthropic a 'Once In a Lifetime' Investment, Says Sameer Dholakia
Sameer Dholakia, Ed Ludlow, Samir
Fireside Chat, Interview, Other
Investment Performance and Current Positioning
- The firm executed its largest-ever investment in Anthropic's history, totaling over $840 million across five funding rounds.
- The initial investment occurred during Series D when Anthropic was valued at $18 billion, contrasting with OpenAI's then-$2 billion valuation.
- The investment was characterized by "ecstatic" returns and is described as the best return in the firm's history.
- The firm held a contentious debate prior to the investment due to Anthropic's anemic gross margins, staggering losses, and lack of backing from major tech giants like Microsoft or Google.
- Strategic conviction was based on the foresight that training AI models requires massive capital expenditure, a pattern previously observed in the cloud computing sector.
- The firm's partners cited their specific backgrounds in cloud computing and enterprise building as the basis for this contrarian view.
- Anthropic's compute model has since evolved from relying on third parties to a diversified "Anthropic Direct" strategy.
- This includes direct leasing of data centers and chips to ensure resilience and fungibility across four different accelerator platforms.
- Following the recent IPO, the firm remains a significant buyer and intends to maintain a long position.
- The firm acknowledges the responsibility to provide DPI (Distributions to Paid-In Capital) to limited partners but will realize gains "in due course" rather than immediately.
- They believe Anthropic's capital structure positions them to be opportunistic across volatile market conditions.
Operational Strategy and Capital Allocation
- Anthropic CFO Krishna Rao has successfully shifted the engineering team's focus toward extreme compute efficiency.
- This internal cultural shift resulted in surprisingly high margins and periods of positive free cash flow in the current year.
- The firm notes that Anthropic's decision to support a multi-chip set (AMD, NVIDIA, Tranium) from inception was a highly contrarian but ultimately superior strategic move.
- This approach provides a diversified supply chain and leverage compared to competitors relying on a single vendor.
- Anthropoc's financial posture is bolstered by a pre-offering balance sheet with over $130 billion in cash, recently raised to nearly $120 billion via the current offering.
- The company maintains a philosophy of being "forward leaning on compute" while remaining conservative on the balance sheet.
- The firm is actively acquiring compute capacity and is positioned to acquire distressed cloud assets or data center assets if market conditions allow.
Broader Investment Thesis and Future Opportunities
- The firm anticipates a "cone of uncertainty" regarding future model capabilities and revenue demand, leading to a diversified investment strategy across the entire AI stack.
- The strategy mirrors the cloud computing era: a handful of dominant model providers will coexist with numerous enterprise-specific application layers.
- The firm is specifically targeting the "last mile" of AI adoption for vertical markets, including law firms, investment banks, healthcare, accounting, and property management.
- Specific portfolio additions and focuses include:
- Fireworks AI: Invested in the final round at a $15 billion valuation; founded by former Meta employees; focuses on open-weight models and inference serving.
- Vertical AI Applications: Active investments in Legora (legal), EvenUp (legal), Abridge (healthcare), Field Guide (accounting), and Elise AI (property management).
- Physical AI: The firm invested $200 million in Waymo early this year, viewing autonomous ride-hailing as a potential trillion-dollar opportunity.
- The firm believes the next generation of "juggernaut" businesses will be founded by employees who gained expertise at companies like Facebook or Anthropic.
- They expect a wave of new startups emerging from the talent pool released by major AI companies.
- While Anthropic's closed model is dominant, the firm maintains a diversified portfolio to hedge against uncertainty, including investments in open-weight model generators like Fal AI.