• Treat AI as a long-term growth theme, not a reason to buy at any price; favor businesses with durable customer adoption, credible growth, and the capital to meet compute needs.
• The strongest private-market themes discussed are frontier-model providers, open-model infrastructure, and legal AI, but no specific public-market buy recommendation or attractive entry price was established.
• Approach Fireworks AI’s potential 20× return claim and Waymo’s trillion-dollar scenario as highly optimistic possibilities, not forecasts; both depend on winning customers and expanding in competitive markets.
• For private investments, scrutinize valuation, retention, margins, liquidity, and fees; the discussion suggests a possible IPO window in late 2027–2028, but timing is uncertain.

Artificial Intelligence and Venture-Capital Themes

• The guest sees AI as a potentially larger market than oil, arguing that demand for “intelligence” could span consumers, businesses, and governments.
• He believes AI is creating unusually rapid company growth and value creation, but also emphasized that entry price still matters. His approach is to assess whether a company can grow into its valuation, with particular attention to the next 12 months.
• He cited 10–15× next-twelve-month revenue as potentially reasonable for exceptionally fast-growing businesses, not as a universal valuation rule.
• Compute requirements and scarce, expensive infrastructure can make AI businesses highly capital-intensive. The guest views access to capital as a competitive advantage, but said funding alone does not guarantee success: product quality and customer adoption remain essential.
• He expects both closed, frontier models and open models to support large businesses. He cautioned against focusing too much on predicting their precise market share.
• AI venture investing carries significant risks, including high entry valuations, capital needs, competition, and uncertainty about which companies will become durable leaders.

Takeaways

• Treat the AI opportunity as a broad theme, not a reason to buy every AI-related company at any price.
• Consider whether a company can sustain growth, build customer loyalty, and fund its compute needs—not just whether it has a compelling AI narrative.
• The discussion concerns mostly private venture investments; it does not provide a specific recommendation to buy publicly traded securities.

Anthropic

• Bessemer invested in Anthropic’s Series D, at a reported $15 billion pre-money valuation and $18 billion post-money valuation.
• The investment thesis was that building frontier models would require substantial compute and capital, potentially concentrating the market among a small number of well-funded providers.
• The guest said Anthropic’s relationship with AWS was important for both compute and distribution.
• He believes frontier AI can address problems that remain unsolved, and said he wished Bessemer had invested more in that round.
• The guest said the investment’s valuation looked modest in hindsight, while acknowledging that the opportunity appeared risky and non-obvious at the time.

Takeaways

• The discussion supports a long-term thesis on frontier AI, but emphasizes how difficult it is to judge early-stage opportunities and appropriate valuations.
• For investors, the relevant considerations include access to compute, distribution partnerships, capital requirements, and the company’s ability to keep improving its models.

Fireworks AI

• The guest described Fireworks as a company in which he is invested and a believer in the open-model opportunity.
• He said Fireworks is focused on providing infrastructure for open models and mentioned that he believes it can win against competitors such as Base10 and Modal.
• His broader thesis is that some enterprises will want to run or control their own AI models rather than rely entirely on outside providers.
• The episode refers to a Fireworks investment price of about $15 billion in a discussion of growth-stage opportunities; the guest said he sees the potential for a 20× return from that level. This was his personal view, not a guaranteed outcome.
• Risks discussed around the category include capital intensity, intense competition, and uncertainty over how open and closed models will divide the market.

Takeaways

• Fireworks illustrates the investment case for infrastructure that helps companies use and manage open models.
• The guest’s return expectation is highly optimistic. A general investor should weigh it against the company’s valuation, competitive position, and ability to convert enterprise interest into durable revenue.

fal (referred to as “Fowl” or “Fall” in the transcript)

• The guest said Bessemer is investing in fal, which he described as being on the generative-media side of the AI market.
• He grouped fal with Fireworks as an example of a business benefiting from greater-than-expected demand for open-model and AI infrastructure.

Takeaways

• The discussion points to generative media as an area of venture interest, but provides no specific valuation, financial details, or return estimate for fal.
• Any investment assessment would need to examine customer demand, costs, competition, and the company’s ability to monetize its tools.

Legora

• Bessemer led an investment in Legora at a reported $1.8 billion valuation, when the company had about $20 million in ARR.
• The guest argued that legal AI could address a large market: he described global legal-services spending as roughly $1 trillion and suggested that legal work currently priced beyond many customers’ reach may create additional demand.
• He said Legora has global ambitions and suggested the legal-services market outside the U.S. could substantially expand the opportunity.
• He prefers businesses with sticky enterprise customers and strong retention, but acknowledged that gross margins in AI businesses may be lower early on and require careful scrutiny.

Takeaways

• The investment case rests on a large potential market, new uses for legal services, and the possibility of scaling beyond the U.S.
• The $1.8 billion valuation relative to current ARR underscores the importance of testing whether future growth can justify a high entry price.
• Watch customer retention, revenue quality, margins, and the company’s ability to turn AI adoption into lasting legal-workflow usage.

Harvey

• Harvey was described as one of the leading companies in legal AI, alongside Legora.
• The guest said that significant funding can help leading companies build brand, attract talent, scale quickly, and reinforce customer adoption.
• He believes both Harvey and Legora can do well, despite their competition.

Takeaways

• The discussion presents legal AI as a competitive market in which strong products and customer adoption matter alongside financing.
• The guest gave no specific valuation or price target for Harvey, so the episode does not establish whether its shares or private-market valuation are attractive.

Abridge

• Abridge was cited as an example of a company applying AI to doctors and healthcare.
• The guest referred to it as a business whose gross margins were lower at an earlier stage than they can be at scale.

Takeaways

• Abridge illustrates the potential for AI to support specialized professional workflows.
• The discussion did not provide a valuation, financial data, or a specific investment recommendation. Margin improvement should be evaluated rather than assumed.

EvenUp

• EvenUp was mentioned among legal-related AI companies in the guest’s portfolio and as an example of a specialized application business.

Takeaways

• The episode supports interest in AI applications for specific professional tasks, but offers no valuation or company-specific investment analysis for EvenUp.
• Assess whether the product is deeply embedded in customers’ work and whether revenue retention supports a durable business.

Fieldguide

• Fieldguide provides AI tools for auditors. The guest said Bessemer invested when it was below $10 million in ARR and at a valuation below $200 million.
• He said the company was growing quickly and that adoption improved after a recent advance in AI models made its solution more compelling.
• He expressed regret that Bessemer had not yet invested more, while distinguishing Fieldguide from companies he views as potentially much larger.

Takeaways

• Fieldguide illustrates how advances in foundation models can make a specialized application more useful and increase adoption.
• The guest’s enthusiasm was not a specific return forecast. Investors should still assess the company’s growth, customer retention, and ultimate market size.

Elise AI

• The guest said Bessemer planned to invest nearly $200 million in Elise AI, describing it as an example of backing companies the firm believes are gaining traction in large markets.
• The transcript provides little company-specific detail beyond the investment and the firm’s confidence.

Takeaways

• The size of the planned investment reflects Bessemer’s conviction, not independent evidence that the opportunity is suitable for all investors.
• The episode does not provide a valuation, revenue figures, or a return estimate for Elise AI.

OpenAI

• The guest said he had passed on an earlier opportunity to invest in OpenAI.
• He described the experience as a reminder that venture investors can miss major opportunities and that their expectations for how large AI companies could become had initially been too limited.

Takeaways

• OpenAI serves in the discussion as a lesson about uncertainty and the risk of underestimating transformative markets.
• The episode does not offer a current valuation analysis or investment recommendation for OpenAI.

Jasper

• The guest recalled seeing an early demonstration of Jasper, an AI marketing application built on GPT-3, in late 2021.
• He described the experience as an early signal that generative AI could create a major technology cycle.

Takeaways

• Jasper was discussed as an example of an early AI application, not as a current investment recommendation.
• The episode gives no current company outlook, valuation, or financial information.

Thinking Machines

• The guest referred to Thinking Machines at a reported $40 billion valuation while discussing the possibility of a large U.S.-based open-model business.
• He said he could imagine a substantial open-model company, but did not say that Thinking Machines itself would reach a particular valuation.

Takeaways

• The discussion supports the possibility of large businesses built around open models, while leaving the economics and eventual scale uncertain.
• A high private valuation should be assessed against growth, monetization, and competition—not just the size of the AI market.

Sarvam

• Bessemer was described as an investor in Sarvam, an Indian AI company.
• The guest highlighted India’s many local languages and the opportunity to train models on local content, including voice and text.

Takeaways

• The thesis is that local language needs and locally relevant data may create a distinct market for regional AI models.
• The discussion provided no valuation or financial information, so it does not establish whether Sarvam is attractively priced.

Mistral

• The guest said Bessemer was investing in, or in the process of investing in, Mistral.
• He used Mistral as an example of a sovereign-model opportunity: some governments and regulated industries may prefer AI models under local or regional control.

Takeaways

• Sovereign AI may be a relevant niche where buyers prioritize control, regulation, or geopolitical considerations.
• The guest’s thesis concerns selected use cases and customers, not necessarily all AI demand. No valuation or price target was provided.

Waymo

• The guest said Bessemer had invested nearly $200 million in Waymo.
• He sees potential beyond ride-hailing, including autonomy for personal vehicles, trucking, logistics, construction, mining, and agriculture.
• He argued that Waymo could become a major business by licensing or extending its autonomy technology to other vehicle uses, and described it as having the potential to become a trillion-dollar business. This was an optimistic personal forecast.
• The episode compares a Waymo valuation of approximately $115 billion with Uber’s market capitalization of approximately $120–125 billion, arguing that ride-hailing is not the only market relevant to Waymo’s potential.
• The guest said Waymo’s immediate focus was ride-hailing and that it was expanding into more cities.

Takeaways

• The investment thesis depends on autonomy expanding beyond ride-hailing into other vehicles and industries.
• The trillion-dollar view is a scenario, not a price target or assured outcome. Key questions include whether Waymo can expand safely and commercially, and whether its technology can serve markets beyond its current focus.

Rocket Lab

• Rocket Lab was mentioned as an example of an investment that David Cowan pursued when others thought the opportunity was unusual; the guest said it later became a very large business.

Takeaways

• Rocket Lab was cited as an example of the potential rewards—and uncertainty—of investing in less obvious opportunities.
• The episode provides no current valuation, forecast, or specific recommendation for the company.

Stripe

• The guest said Bessemer regretted not investing in Stripe, particularly given the firm’s history of investing in developer-focused companies such as Twilio and SendGrid.
• Stripe was also discussed as an example of a successful company remaining private, and of the strategic flexibility that private companies may have.

Takeaways

• The example illustrates the opportunity cost of missing a strong private company, but does not provide a current valuation or investment recommendation.
• The broader point is that companies may stay private longer, meaning some growth may occur before public-market investors can participate.

Amazon (AMZN), Microsoft (MSFT), and Alphabet (GOOGL)

• The guest used the cloud industry as an analogy for frontier AI: a capital-intensive market may consolidate around a small number of large providers.
• He cited Microsoft’s partnership with OpenAI, Google’s Gemini effort, and AWS’s investment in Anthropic as examples of how major technology companies are positioned in AI.
• He described the infrastructure market as requiring substantial compute and capital, while expecting many applications to be built on top of the large platforms.

Takeaways

• The discussion suggests that large technology companies may have advantages in funding, compute, and distribution.
• It also distinguishes infrastructure providers from the many application companies that could be built on their platforms. The episode provides no stock-specific valuation analysis or recommendation.

Tesla (TSLA) and Uber (UBER)

• Tesla was mentioned in connection with its Full Self-Driving offering as an example of autonomy technology being applied to privately owned cars.
• Uber was used as a market-cap comparison for Waymo; the guest argued that comparing the two only on ride-hailing misses Waymo’s potential applications in other markets.

Takeaways

• The discussion presents autonomy as a potential growth theme, but does not provide a recommendation on Tesla or Uber.
• The Waymo comparison depends on future expansion into areas beyond ride-hailing; it should not be treated as a direct valuation comparison between the companies.

Twilio and SendGrid

• The guest used Twilio and SendGrid as examples from the earlier cloud and developer-software era.
• He said that once developers had chosen certain communication APIs, those providers could become the default choice, creating a reinforcing adoption advantage.
• He compared this dynamic with the possibility that leading AI companies could gain advantages through capital, product quality, brand, and customer adoption.

Takeaways

• The historical comparison highlights how developer adoption and customer habit can create durable advantages.
• It does not establish that today’s AI leaders will follow the same path; competition and capital requirements are different.

Private Venture Capital

• Bessemer announced funds totaling $5.75 billion, including a $1.75 billion seed and early-stage fund and a $4 billion growth fund, according to the discussion.
• The guest argued that higher early-stage valuations and capital needs can make it harder for small funds to build successful portfolios.
• He believes venture firms may need to be either large, global platforms or highly specialized investors to compete effectively.
• Bessemer’s growth strategy includes concentrated positions; the guest said a late-stage growth investment could represent 5–10% of a fund, and noted that one prior fund had 20% in Anthropic.
• He said early-stage funds need the possibility of very large winners, while growth funds can target lower but still substantial multiples. He described 5×-plus fund returns as the firm’s goal and said early-stage investments may need 50× outcomes to compensate for failures.
• He expects companies to remain private longer, but predicted that IPO activity could improve in late 2027 and 2028, citing potential public-market demand for AI exposure.
• He also said venture investors should seek opportunities to return capital to their limited partners, including through late-stage share sales.

Takeaways

• Private venture exposure is a high-risk, illiquid investment category; the guest’s discussion is about fund strategy, not a general recommendation to allocate to venture capital.
• Large funds, concentrated bets, and high return targets do not remove the risk of loss or guarantee access to top-performing investments.
• Investors considering private-market exposure should weigh liquidity, fees, diversification, manager selection, and the possibility that strong performance may be concentrated in a small number of companies.