• NVIDIA (NVDA) stands out as the panel’s preferred AI exposure because its differentiated technology is more essential than capacity providers’ infrastructure.
• Be cautious with CoreWeave (CRWV) and Oracle (ORCL): their capital-intensive AI buildouts and debt could become a burden if demand growth disappoints.
• Treat Micron (MU)’s reported 350% one-year gain as historical, not a forecast; the panel noted meaningful risk in secondary AI beneficiaries.
• For any potential Anthropic IPO, assess growth, market share, and valuation carefully—the discussed $1.5 trillion figure is not a confirmed price or buy signal.

NVIDIA (NVDA)

• NVIDIA was described as the leading, “A-tier” company in the AI infrastructure ecosystem, with core technology that other providers depend on.
• The discussion contrasted NVIDIA with infrastructure providers such as CoreWeave and Oracle, arguing that those businesses are more exposed to changes in AI demand.

Takeaways

• The panel’s comparison favors businesses with differentiated, essential technology over providers whose position depends more heavily on supplying capacity.
• This was a qualitative view, not a valuation analysis or a specific recommendation to buy NVIDIA.

CoreWeave (CRWV)

• CoreWeave’s shares were said to have fallen roughly 15%–20% during a difficult week; another speaker referred to a decline of about 25%.
• The panel described CoreWeave as a riskier AI infrastructure provider because it must fund rapid growth with substantial capital expenditure and debt. One speaker cited roughly $7 billion in run-rate revenue, $10 billion–$20 billion in capital expenditure, and $35 billion–$40 billion in debt, while noting the figures were from memory.
• The business depends on continued demand for computing capacity from large customers such as Microsoft and OpenAI. If demand grows more slowly than expected, the company may be left with expensive capacity and debt obligations.
• One speaker argued CoreWeave is a “B-tier” asset compared with NVIDIA because it does not have a uniquely differentiated chip and software platform. Another cautioned that the share-price decline could prove excessive if end-user demand remains strong.

Takeaways

• The investment case discussed is highly dependent on AI demand growing fast enough to absorb ongoing infrastructure spending.
• For investors considering the stock, the panel’s central risk check is whether expected demand can keep pace with CoreWeave’s capital spending and debt burden.

Oracle (ORCL)

• Oracle was also characterized as a “B-tier” AI infrastructure asset relative to NVIDIA.
• The panel said Oracle is renting out excess capacity, buying GPUs, and participating in the Stargate initiative, while arguing that its traditional enterprise software business is not selling enough to explain that infrastructure push.
• Oracle was grouped with highly leveraged infrastructure providers that could react sharply to any signs of slowing AI demand.

Takeaways

• The discussion suggests separating Oracle’s established software business from its newer, capital-intensive AI infrastructure ambitions.
• The risk highlighted was exposure to expensive capacity investments if demand or customer growth disappoints.

Micron (MU)

• Micron was cited as an example of a second-tier technology stock that had performed exceptionally well: one speaker said it was up about 350% over the prior year.
• The same discussion acknowledged the risks of holding such stocks and did not present Micron as a low-risk alternative to NVIDIA.

Takeaways

• The example illustrates that secondary beneficiaries of an AI or semiconductor boom can deliver substantial gains, but the panel also stressed that those gains come with meaningful risk.
• The cited return is historical performance from the discussion, not a forecast.

Anthropic (Private; potential IPO discussed)

• Anthropic was discussed as a possible IPO candidate, with a potential valuation reportedly shifting from $2 trillion to $1.5 trillion.
• The panel cited $16.8 billion in Q3 revenue, up 45% from Q2, and an 85% gross margin. It also discussed the company becoming profitable on an adjusted basis.
• Speakers saw strong absolute growth, but warned that growth rates may slow from exceptionally high levels and that developer preferences and AI market share can change quickly.
• They noted that market-share shifts could lead to price competition among leading AI providers. One speaker said Anthropic’s share could decline even while its revenue continued to grow.
• The discussion also highlighted “code review” and the broader process of getting AI-generated code into production as important opportunities, but warned that foundation-model companies could compete directly with specialist software providers.

Takeaways

• The potential investment case rests on continued revenue growth and sustained demand for AI products, not just the headline valuation.
• The risks specifically raised were slowing growth, shifting market share, and future price competition. A potential IPO valuation is not a price target or assurance of investment returns.

OpenAI (Private)

• OpenAI was described as one of the largest sources of demand for AI computing, alongside Anthropic.
• The panel suggested that combined growth among the largest AI model providers is an important indicator of the broader AI economy and, indirectly, of demand for data-center capacity.
• OpenAI was also discussed in connection with potential future IPO activity, but no confirmed timing or valuation was established in the conversation.

Takeaways

• The discussion treats OpenAI’s growth as a demand signal for the wider AI infrastructure market.
• Investors following infrastructure companies may want to distinguish durable growth in end-user demand from shifts in market share between OpenAI and competitors.

Modal (Private)

• Modal raised $750 million in a Series D at a $15.75 billion valuation. The discussion said its run-rate revenue had surpassed $1 billion and had accelerated sharply from less than roughly $100 million in the prior year.
• Speakers viewed the round positively because Modal provides inference services and developer-focused products, rather than simply supplying GPU capacity.
• The panel saw a large potential market for independent inference providers, especially if businesses can obtain comparable model output at lower cost than buying directly from model developers.
• Risks mentioned included heavy competition and the possibility that infrastructure providers could move up the stack and offer inference services themselves. The panel also discussed margins as an important factor.

Takeaways

• The opportunity discussed is the growth of the inference market and the value of making AI models cheaper and easier for developers to use.
• The key questions are whether Modal can sustain its growth, defend its position amid competition, and improve or maintain margins at a high valuation.

AI Code Review and “Build-to-Production” Software

• The panel called code review—and, more broadly, moving AI-generated software from development into production—one of the most important software opportunities.
• Greptile and CodeRabbit were mentioned as specialist companies in the category. The discussion also noted that foundation-model providers such as Anthropic could compete directly with them.
• Speakers argued that AI can generate code quickly, but teams still need to test, review, and make that code production-ready. They described this broader workflow as potentially more important than code generation alone.
• Replit was used as an example of a company adjusting which AI models it routes users to. The discussion said model quality can be close enough that lower cost becomes decisive, particularly when customers are sensitive to variable AI bills.

Takeaways

• The investment theme is not limited to code generation: testing, review, deployment, and production reliability may also benefit as more software is created with AI.
• The risk is that fast-growing standalone applications may be vulnerable to competition from the foundation-model companies that enable them.

Anduril (Private)

• Anduril’s valuation was discussed at about $90 billion, with the conversation questioning whether that represented roughly 40 times revenue.
• Speakers described the company as a promising defense technology business with contracts and a strong position in next-generation defense.
• One speaker argued that a high valuation could be supported if revenue continues to grow rapidly and the company maintains a strong pipeline of government business. The discussion also warned that this requires substantial future growth to justify the price.

Takeaways

• The panel’s debate was less about whether Anduril is a strong company and more about whether its valuation already assumes too much future success.
• The relevant risks raised were the high revenue multiple and the need for continued growth and contract visibility.

Town (Private)

• Town raised $108 million at a $1 billion valuation.
• One speaker described it as an early-stage bet with a smart team and some interesting revenue, but said there was not enough information to form a strong investment view.
• The panel compared a billion-dollar valuation to the “new” benchmark for a conventional Series B, while warning against treating that figure as small or routine.
• The discussion also referred to a potential $500 million need for inference spending over the next 12 months, though the context was conversational and not fully explained.

Takeaways

• The round illustrates how high private-market entry valuations have become, even for early-stage companies.
• The panel’s view was cautious: a promising team and large market opportunity are not substitutes for evidence about revenue, costs, and the amount of capital the company will need.

Bending Spoons (Private)

• Bending Spoons was described as a holding company that buys software businesses and aims to improve and hold them over time, rather than quickly reselling them.
• The discussion said it had acquired assets such as Vimeo, Evernote, and Eventbrite, and argued that pricing changes and operational improvements could make some acquired products profitable.
• The panel compared its approach to Berkshire Hathaway’s long-term ownership model, while also questioning whether software products can remain durable for as long as consumer staples such as See’s Candies.
• Berkshire Hathaway was said to have bought $330 million of new shares, with the investment reportedly boosting Bending Spoons’ share price.

Takeaways

• The investment theme is long-term ownership and operational improvement of established software assets, rather than betting only on high-growth startups.
• The key uncertainty is whether Bending Spoons can keep acquired software products relevant and valuable as technology and customer needs change.

Berkshire Hathaway (BRK.A / BRK.B)

• Berkshire’s reported investment in Bending Spoons was presented as a vote of confidence in the company’s long-term software ownership strategy.
• The panel’s comparison to Berkshire’s holdings was conceptual: the idea was to acquire durable businesses and hold them, not that software necessarily has the same longevity as Berkshire’s best-known consumer businesses.

Takeaways

• The discussion highlights a possible institutional interest in software businesses that generate cash and can be improved over time.
• Berkshire’s reported investment is a signal to consider, not proof that the investment will succeed or that the same strategy will work across software assets.

Tesla (TSLA) and SpaceX (Private)

• Tesla’s new Roadster and Elon Musk’s marketing were discussed, with speakers arguing that Musk’s public profile can support a valuation premium.
• One speaker predicted that Tesla and SpaceX could be merged soon, describing the Roadster as a possible final major event for Tesla as a standalone company. This was expressed as speculation, not a confirmed transaction.
• The panel also argued that China could dominate global electric vehicles and autonomy, while suggesting that autonomous vehicles could reduce the need for people to drive themselves.
• SpaceX was discussed as a major source of venture returns. Harvard’s endowment was said to hold SpaceX shares valued at about $2.2 billion, and the panel noted that a few large private-company outcomes can have a major effect on institutional returns.
• A proposed distribution of SpaceX shares to children was discussed as a possible marketing and wealth-distribution idea; it was not presented as an established investment opportunity.

Takeaways

• For Tesla, the discussion points to potential support from brand and founder-driven enthusiasm, alongside uncertainty about competition and the company’s future structure.
• SpaceX was cited as an example of how private-company gains can drive venture and endowment performance, but those gains are not necessarily accessible to ordinary investors at the same terms.

General Atlantic (Potential IPO)

• The panel discussed General Atlantic’s potential IPO and the economics required for a large private-markets firm to go public.
• One speaker estimated that a firm might need roughly $800 million to $1 billion in fee income to reach the scale required for a meaningful public listing.
• The discussion noted that public investors tend to value recurring management fees more readily than performance fees, or carry.
• Speakers also warned that public-market scrutiny and pressure to grow assets could make a firm more risk-averse and change its culture.

Takeaways

• The potential opportunity depends on recurring fee income, scale, and continued growth—not just the firm’s investment reputation.
• The risks raised were public-market pressure to grow and the possibility that becoming public could change how the firm operates.

Public Software Valuations: Braze (BRZE), monday.com (MNDY), HubSpot (HUBS), Klaviyo (KVYO)

• The panel used public software companies as a comparison for private-company valuations. It cited roughly 2 times revenue for monday.com, HubSpot, and Klaviyo, and said Braze had been trading at a higher multiple before a reported deal valued at about $5.8 billion.
• Speakers argued that some private AI-related companies may be priced far above public software comparables, and questioned whether those premiums will hold if growth slows.
• They also cautioned that AI-enabled companies may not automatically receive higher multiples than older software businesses once their growth rates moderate.

Takeaways

• Comparing private valuations with public-company multiples can help investors assess how much future growth is already priced in.
• The discussion’s main warning was that high growth may justify premium valuations, but a slowdown could expose companies to significant multiple compression.

Bootstrapped and Profitable Software: 37signals and Zapier

• The panel discussed 37signals as an example of a profitable, relatively independent software business. A claim that it generated $370 million in profit since 2020 was challenged during the conversation, so the figure should be treated as uncertain.
• Zapier was described as a valuable company with about $300 million in revenue, but with growth that may have slowed and an IPO not viewed as likely in its current form.
• Speakers debated whether raising a small amount of venture capital and then growing independently—a “one-and-done” approach—can be less attractive than raising no venture capital. They noted that distributing profits or buying out investors can be difficult.

Takeaways

• Profitable, bootstrapped software businesses can create substantial value without pursuing the largest possible venture outcome.
• Founders and investors should consider the trade-offs between venture funding, ownership, cash distributions, and the practical difficulty of buying investors out later.

Venture Capital and Private-Market Funding

• The panel described venture funding as increasingly concentrated in a small number of companies and categories with very large potential outcomes.
• Some businesses with strong growth and profitability may struggle to raise capital if investors fear that AI will disrupt their market. Meanwhile, areas such as data centers, space, and quantum computing may still attract funding because of their perceived upside.
• One speaker advised founders outside the most fashionable categories to meet a very large number of investors—potentially 100—rather than expecting a term sheet based on strong metrics alone.
• The panel also argued that large venture funds need investments capable of producing very large returns, which can make smaller outcomes less attractive to those funds.
• Early-stage investors may face high entry valuations and reduced allocations when larger funds follow on.

Takeaways

• For founders, the practical lesson from the discussion is to expect a selective fundraising market and broaden the investor outreach process.
• For investors, the panel highlighted the trade-off between chasing a small number of enormous outcomes and backing solid companies at more moderate valuations.

Venture-Backed Companies and Endowment Returns

• Harvard’s endowment was reported to have returned 27.1% on a $68 billion portfolio, with venture investments contributing to performance.
• The panel pointed to SpaceX as an important holding and said large private-company outcomes can materially affect endowment returns.
• Speakers speculated that proceeds from major private-company investments could flow back into venture funds, potentially making the market more competitive. This was a forecast, not a confirmed trend.

Takeaways

• Institutional returns can be strongly affected by a small number of private-company investments, which makes those outcomes difficult to reproduce consistently.
• The discussion suggests watching for renewed capital flows into venture, while avoiding the assumption that past gains in a few standout companies will be repeated.