• Consider CoreWeave (CRWV), Nebius, and Crusoe for exposure to tight GPU capacity demand, but prioritize operators that can bring clusters online reliably and avoid locking in too much capacity at below-market prices.
• Monitor CoreWeave’s shift toward spot-market exposure; rising rental prices and customer bookings extending into 2027 support the opportunity, while deployment delays and contract terms remain key risks.
• Nvidia (NVDA) may benefit if its DSX Pod helps customers deploy GPUs faster, though the discussion provided no price target or explicit buy recommendation.

Anthropic (Private)

• Gavin Baker described Anthropic’s S-1 as an upside surprise: reported cash burn, profitability, cash generation, and unit economics were encouraging.
• He said Anthropic had reached roughly half of the Fortune 500, while its products still represented a small share of those customers’ expenses—evidence, in his view, that enterprise AI adoption remains early.
• The S-1 showed a more diversified customer base than some reports had suggested, including 11,000–12,000 accounts spending over $100,000 and 250 spending over $10 million.
• Baker said Anthropic had reaccelerated and was gaining share in October, after losing share earlier in the summer. He also noted that 74% of its compute was devoted to training—a potential strength if it produces better models, but a risk if training needs fall and that capacity shifts into inference.

Takeaways

• The discussion supports a bullish view of Anthropic’s growth and economics, while highlighting that enterprise adoption may have further room to expand.
• Watch whether customer growth and usage broaden, and whether heavy training investment translates into durable revenue. No public ticker or price target was mentioned.

OpenAI (Private)

• Baker said OpenAI accelerated during the summer, alongside growth in open-weight model providers.
• The hosts discussed the possibility of OpenAI eventually becoming public, which Baker said could affect how much capital mutual funds can allocate to private-market investments.

Takeaways

• OpenAI was presented as a major AI competitor benefiting from strong demand, but the conversation offered no specific financial figures or investment recommendation.
• Any public-market investment opportunity remains contingent on a future listing; none was announced in the transcript.

SpaceX (Private)

• SpaceX was cited as an example of a capable operator able to bring GPU capacity online and run it effectively.
• Baker said SpaceX has been insourcing compute previously associated with Cursor, and the hosts discussed a post-merger integration involving SpaceX and xAI.
• Baker suggested that companies selling compute into the spot market may benefit from rising prices and short payback periods, rather than locking in long-term contracts at fixed rates.
• He also said SpaceX going public could unlock buying capacity in private markets for mutual funds. The hosts discussed Elon Musk’s plan to give every child in America a share of SpaceX.

Takeaways

• The discussion frames SpaceX as a potential AI infrastructure and model-distribution player, as well as a possible future source of publicly traded shares.
• These are private-company prospects, not current stock recommendations; no IPO date or valuation was given.

xAI and GrokBot (Private)

• Baker said GrokBot was the most impactful AI product for him and his firm so far, praising its ease of use and usefulness in delivering information to the right people.
• The hosts discussed Cursor’s product and distribution strengths as part of the potential go-to-market opportunity for GrokBot.

Takeaways

• The investment theme is that useful AI agents and strong distribution may matter as much as model benchmark performance.
• The transcript provides positive user feedback, but no revenue figures, valuation, or specific investment recommendation.

Nvidia (NVDA)

• Baker highlighted Nvidia’s DSX Pod, a pre-assembled, seven-rack system intended to make GPU deployments easier.
• He argued that faster installation could reduce the time and working capital required to get GPUs energized and generating revenue.

Takeaways

• The discussion points to a potential benefit for Nvidia if its systems simplify data-center deployment and help customers bring capacity online faster.
• The transcript did not give a price target or a specific recommendation on Nvidia shares.

GPU Cloud Providers: CoreWeave (CRWV), Nebius, and Crusoe

• Baker said demand for GPU compute had tightened: spot rental prices were rising, and customers were paying more for future capacity, including clusters planned for 2027.
• He cited Silicon Data’s assessment that the compute forward market had moved into contango, meaning future capacity was priced above spot capacity.
• Baker praised CoreWeave’s execution and reliability but said it had locked in substantial compute through long-term contracts at prices now well below market.
• He said CoreWeave was seeking more spot-market exposure and fewer long-term offtake agreements. He also named Nebius and Crusoe among operators with a track record of bringing GPUs online.
• Baker argued that financing conditions may favor experienced operators that can pre-sell capacity or obtain customer prepayments.

Takeaways

• Tight compute supply and rising forward prices could support revenue growth for capable GPU cloud operators.
• Operator execution and contract structure are important risks: long-term fixed-price agreements may leave providers selling capacity below current market rates, while delays in energizing GPUs can tie up capital.
• The transcript was positive on CoreWeave’s operations but critical of its contract pricing; it did not provide a stock price target.

Open-Weight Inference Providers: Fireworks, Base10, Together, and Modal

• Baker said open-weight-focused companies—including Fireworks, Base10, Together, and Modal—were growing faster than OpenAI or Anthropic at the time discussed.
• Atreides led Fireworks’ most recent funding round. Baker said Fireworks was growing so quickly that its valuation was difficult to settle, with revenue increasing between the initial term sheet and the round’s close.
• The hosts also mentioned Cognition as part of the broader group of fast-growing AI companies.

Takeaways

• The discussion suggests that AI investment opportunities may extend beyond model developers to companies providing inference infrastructure and services.
• These businesses are private, and the growth comments are not accompanied by audited figures or valuation details.

Google (Alphabet: GOOGL, GOOG) and Meta (META)

• Google was mentioned as one of the providers some companies may be choosing as they adopt AI; the broader discussion emphasized that many large companies have not yet worked through enterprise-wide deployment.
• Meta appeared in a discussion about large AI customers and potential spending concentration. Baker said Anthropic’s S-1 showed a more diversified customer base than some claims that a small number of customers drove most of its revenue.

Takeaways

• The transcript frames competition among major AI providers and uneven corporate adoption as relevant themes for Alphabet and Meta.
• It did not make a direct bullish or bearish recommendation on either company or provide company-specific targets.

AI Infrastructure and Compute Demand

• Baker remained bullish on the ability of the market to absorb new GPU supply, invoking the idea that greater efficiency can stimulate greater usage.
• He said that, despite substantial new capacity expected, companies were still struggling to secure compute well into 2027.
• He emphasized that supplying GPUs is not enough: operators must also build clusters, connect power, and run them reliably. Some deployments can take months to become productive.
• The discussion suggested that cloud revenue and operating cash flow could keep accelerating, potentially financing more of the 2027 build-out internally.

Takeaways

• The investment theme is sustained demand for AI compute, alongside the infrastructure needed to power and operate it.
• The key risks mentioned were deployment delays, operator execution, and the possibility that expected supply growth could eventually outpace demand.

AI in Biology and Drug Development

• Baker said AI has already helped identify potential uses for existing drugs, including a possible treatment for a rare and devastating disease affecting children.
• He described AI-enabled biology as promising but acknowledged that drug development and commercialization may take years.

Takeaways

• AI-driven drug discovery is a long-term opportunity, with potential benefits even from repurposing existing compounds.
• The main risk raised was the time required to turn scientific progress into approved drugs and revenue. No specific company or investment recommendation was named.

Rare-Earth Processing (Private, Company Unnamed)

• Baker said Atreides had invested privately in a company developing a chemical process for refining rare earths.
• He noted that the challenge is not necessarily a lack of rare-earth deposits, but that refining is a dirty process.

Takeaways

• The opportunity discussed is in improving processing technology rather than acquiring mines, potentially addressing an industrial bottleneck.
• The company was not named, and no financial details or specific investment terms were provided.

Gold, Copper, and Asteroid Mining

• The hosts referred to copper as a possible broad proxy for AI infrastructure demand, but Baker did not endorse a specific copper investment.
• Baker said he had historically viewed gold as useful for portfolio construction, while also noting that asteroid mining could eventually challenge assumptions about scarce resources and supply.
• He compared the possibility of technological disruption in gold to the disruption synthetic diamonds brought to the diamond industry.

Takeaways

• The conversation raises two different themes: potential demand for industrial materials such as copper, and the long-term possibility that technology could alter the supply outlook for scarce commodities.
• No commodity price target, timeline, or specific recommendation was given.