• Consider open-weight AI models as a potential momentum theme over the next couple of years, but the discussion named no specific investable company or ticker.
• Treat CoreWeave (CRWV) as a high-risk, credit-sensitive exposure: its bonds reportedly yielded 15–16%, while CDS pricing implied substantial default risk; strong compute demand does not guarantee it can finance its commitments.
• Monitor Bitcoin (BTC) for long-term security-budget risks as block rewards decline; at the cited $82,000, the discussion offered no price target or buy recommendation.
• Zcash (ZEC) and Pearl are speculative alternatives tied to Bitcoin-security and AI-inference themes; consider them only with awareness that neither thesis was supported by a price target or demonstrated investment recommendation.

Bitcoin (BTC)

• Bitcoin and the broader crypto market held up better than the AI-related equity market during a period of sharp volatility.
• The discussion raised concern about Bitcoin’s security budget: miners may shift computing resources to AI inference if it pays better than Bitcoin mining.
• As block rewards decline over time, the speakers questioned whether transaction fees will be sufficient to fund network security. They noted that Bitcoin’s community has historically been reluctant to make changes.
• Bitcoin was cited at $82,000 during the conversation, but no price target or forecast was given.

Takeaways

• The near-term comparison with AI stocks was favorable for Bitcoin, but the speakers identified a longer-term question about how the network will remain secure as mining rewards decline.
• Treat the security-budget concern as a risk to monitor, not as a prediction that Bitcoin will fail. The discussion also emphasized Bitcoin’s established digital-gold narrative.

Zcash (ZEC)

• Zcash was discussed as a potential beneficiary if investors seek alternatives to Bitcoin, including amid concerns about Bitcoin’s future security budget.
• The speakers also connected interest in Zcash to concerns about quantum computing, though they did not provide technical detail or a timeline.
• One speaker described the possibility of flows moving from Bitcoin to Zcash as a bullish case, while another cautioned that Bitcoin remains in a league of its own.

Takeaways

• Zcash was presented as a higher-uncertainty alternative thesis, not as a proven replacement for Bitcoin.
• The discussion offered no price target or specific recommendation. Any investment case depends on whether the concerns cited actually lead investors to shift their holdings.

Pearl

• Pearl was described as a Bitcoin-derived network that aims to combine blockchain security with useful AI computation.
• Instead of miners solving Bitcoin’s energy-intensive puzzles, Pearl is designed to use matrix multiplication for AI inference. Miners could earn payments from inference providers as well as Pearl token emissions.
• The speakers said Pearl had an inference endpoint through Together AI and discussed a Gemma 4 endpoint. They argued that payments for real inference work could make miners less reliant on selling tokens than Bitcoin miners.
• A key uncertainty raised was whether the network can ensure miners are doing genuine, useful inference rather than optimizing for token rewards.

Takeaways

• Pearl is a speculative way to gain exposure to the idea of linking AI inference with blockchain mining.
• Before evaluating the token, investors would need to understand whether there is sustained demand for its inference services, how those services are verified, and whether cash payments meaningfully support the network’s economics. The transcript provides no price target or recommendation.

Anthropic and OpenAI

• The speakers described substantial uncertainty around the economics of frontier AI companies, including high compute and training costs, questions about monetization, and competition from open-weight models.
• Anthropic was said to be losing market share, while both Anthropic and OpenAI were described as continuing to grow collectively. The discussion questioned whether Anthropic could reach a $100 billion run rate by year-end.
• The speakers noted that Anthropic had reportedly become free-cash-flow positive in Q3, but questioned the significance of that claim because it was described using adjusted measures.
• One speaker said they would not buy Anthropic or OpenAI at the valuations being discussed, including a potential Anthropic IPO valuation above $1 trillion. Another said they might consider these companies over a much longer horizon, describing them as potential long-term survivors.
• Enterprise API sales were presented as a potentially attractive business, with one speaker citing estimates of 80–90% gross margins. The discussion also raised questions about customer stickiness and whether customers could shift to cheaper open-source models.

Takeaways

• The transcript presents a mixed outlook: potentially durable long-term businesses, but meaningful uncertainty about near-term growth, spending, and valuation.
• The speakers explicitly warned that AI-related public-market names could remain highly volatile, with 10% daily moves possible. They also cautioned against using leverage.
• A potential IPO valuation above $1 trillion was discussed, but no public-market ticker or specific share-price target was given.

CoreWeave (CRWV)

• CoreWeave was identified as particularly exposed to the growth and financing plans of Anthropic and OpenAI because it relies on large customer commitments.
• The speakers said CoreWeave bonds were trading at yields of roughly 15–16% and that its credit-default swaps at one point implied about a 60% chance of default over the next few years.
• They described a risk that AI infrastructure providers could struggle if customers slow their growth or cannot continue raising capital. One speaker expected painful restructurings across some companies in the ecosystem, while saying they did not expect CoreWeave itself to go bankrupt.
• The discussion also noted that demand for compute may remain strong, but that capital availability and the mismatch between long-lived infrastructure and financing needs are concerns.

Takeaways

• CoreWeave was presented as a high-risk, credit-sensitive AI infrastructure exposure, not a straightforward way to invest in AI growth.
• The cited bond yields and CDS pricing reflect serious market concern, according to the speakers. Investors should distinguish demand for compute from a company’s ability to finance its commitments.

Meta Platforms (META), Amazon (AMZN), Alphabet (GOOGL/GOOG) and NVIDIA (NVDA)

• Meta, Amazon and Google were discussed as major AI infrastructure spenders that may need to tap debt markets after using significant free cash flow.
• The speakers raised concern about whether spending on data centers and AI infrastructure will produce an adequate return, particularly if leading AI companies’ growth slows.
• Jensen Huang of NVIDIA was cited as expecting AI infrastructure spending to rise from roughly $1 trillion to $4 trillion by 2030. The speakers acknowledged that this is a large forecast and that NVIDIA’s perspective may reflect its business interests.
• One speaker argued that NVIDIA’s current hardware may deliver better performance per chip for some inference workloads than open-source alternatives, while the broader discussion emphasized uncertainty about which companies will capture the returns from AI investment.

Takeaways

• The discussion supports monitoring both AI infrastructure spending and the returns companies generate from it; high spending alone does not establish that the investment will pay off.
• The speakers did not give stock-specific price targets or recommendations for these companies.

AI Infrastructure and Open-Weight Models

• The speakers described a possible duration mismatch: AI infrastructure requires large, long-term investment, while demand, customer commitments and financing may be less certain.
• They warned that the AI ecosystem is not uniform. Some companies may be durable, while lower-tier businesses could be wiped out if capital becomes harder to obtain.
• Open-weight models were described as gaining momentum. One speaker said they were “very long” open-weight models in the U.S. and viewed them as a potential momentum trade over the next couple of years.
• The speakers cautioned that open-source models are not necessarily free to run: inference still requires compute, which may benefit infrastructure providers.
• They also noted that older GPUs were retaining value and that some companies were making money brokering them, while questioning whether the broader compute-market businesses being promoted would prove durable.

Takeaways

• The transcript favors looking beyond frontier-model developers to businesses that can use AI to improve productivity or provide compute—while recognizing that financing and business durability remain key risks.
• The stated preference for open-weight models was a speaker’s view, not a quantified recommendation. No specific open-weight-model company or price target was provided.

Character.AI

• Character.AI was cited as having entered bankruptcy.
• The speakers said bankruptcy filings showed the company spending about $9 for every $1 of revenue, even after attempting to use open-weight models.
• The example was used to illustrate concerns about weak unit economics among some AI startups.

Takeaways

• The discussion treats Character.AI as a cautionary example of rapid growth or a popular product not necessarily translating into a viable business.
• No public ticker or specific investment recommendation was mentioned.

Ramp and Anduril

• Ramp was described as a private fintech company with notable AI-related work. One investor reportedly viewed it as their AI exposure.
• The speakers suggested that established businesses able to use AI to reduce costs may be better positioned than companies competing solely to build the most powerful model.
• Ramp’s valuation was said to have reached $60 billion in a fundraising round. Anduril was also mentioned as a company the speaker would classify among stronger, “tier one” businesses, but no valuation or specific investment thesis was provided.

Takeaways

• The discussion favors the idea of investing in businesses that can apply AI within an existing operation, rather than assuming every AI startup will become durable.
• Ramp and Anduril were discussed as private companies; the transcript did not provide a public ticker or a specific recommendation.

NuBank (NU), Revolut and Stablecoins

• NuBank and Revolut were discussed as leading digital banks with potential to benefit from stablecoin use.
• NuBank was described as having 140 million customers. A speaker said stablecoins made up the majority of activity in its crypto wallet, with stablecoins accounting for roughly 75% of wallet volume.
• The discussion highlighted stablecoins as a way for people in countries with weaker currencies to access U.S. dollars. One speaker argued that U.S. stablecoins could help reinforce the dollar’s global role.
• Speakers also discussed the possibility that people may increasingly choose stablecoins or other assets over local currencies. They disagreed about how far this could go, noting that governments have incentives to retain their own currencies and can restrict capital movement.
• The conversation mentioned USDC as an example of a dollar stablecoin but did not compare stablecoin issuers or provide a recommendation.

Takeaways

• Stablecoin adoption and digital banking were presented as potential long-term themes, particularly where customers want access to dollars.
• The opportunity depends partly on continued adoption and on whether local governments allow access. The speakers emphasized that capital controls and monetary sovereignty can limit this trend.

Robinhood (HOOD) and Tokenized Equities

• Robinhood Chain was discussed as part of a broader trend of financial companies building their own crypto infrastructure.
• The speakers noted that activity on Robinhood Chain was still largely memecoin trading, raising a question about the quality and durability of that activity.
• Tokenized equities were presented as a fast-growing access product. The speakers cited an increase in holders from 170,000 at the start of 2025 to 4.7 million, and said tokenized-equity volume had risen roughly 550-fold over the preceding year.
• One speaker said about 65% of tokenized-stock trading on Solana occurred outside U.S. market hours, suggesting demand from users in other time zones.

Takeaways

• The discussion points to rapid growth in tokenized-equity access, especially outside U.S. trading hours, but does not establish how much of that activity is durable or profitable.
• Robinhood Chain’s reported memecoin concentration is a risk to consider when assessing the broader ecosystem. No stock price target or recommendation was given.

Solana (SOL)

• Solana was mentioned as a venue for tokenized equities, with substantial trading reportedly occurring outside U.S. market hours.
• The speakers used the growth in tokenized-equity holders and volume as evidence of increasing interest in on-chain access to traditional assets.

Takeaways

• Tokenized equities were presented as a potential use case supporting activity on Solana, but the transcript did not quantify the resulting revenue or provide a SOL-specific investment recommendation.
• The speakers’ figures suggest rapid growth in usage, but do not by themselves establish long-term economic value for the token.

Hyperliquid (HYPE), Backpack and On-Chain Derivatives

• Hyperliquid was described as having generated interest with a proposal to add options through HIP-4, complementing its existing perpetual-futures activity.
• The speakers noted renewed interest in on-chain options and said Derive Protocol had grown from below $100 million to roughly five times that level over a short period.
• Backpack was discussed as an exchange with a broader ambition to provide shared liquidity and licensed infrastructure across markets.
• A key disagreement concerned U.S. regulation. One speaker argued that U.S. rules could require separate liquidity pools, limiting the promise of a single global shared-liquidity layer. Another emphasized the potential benefits for smaller firms, including improved capital efficiency.
• The speakers also noted that U.S. competition is intense, while overseas markets may offer additional demand.

Takeaways

• On-chain options, derivatives and more efficient liquidity are active investment themes, but regulatory structure could limit how much liquidity can be shared across jurisdictions.
• The transcript identifies product development and market access as potential growth drivers, but offers no token price targets or specific recommendations for Hyperliquid or Backpack.

Polymarket and Prediction Markets

• Prediction markets were described as a prominent topic at Token 2049, alongside growing institutional interest in crypto infrastructure.
• Polymarket was mentioned as one of several businesses exploring global markets and liquidity, but the transcript did not give a company-specific financial outlook or valuation.

Takeaways

• Prediction markets were identified as an active area of interest, but the discussion did not provide enough detail to assess any particular platform’s investment merits.
• Regulatory access and the ability to operate across markets remain relevant considerations in the broader discussion.

OpenSea and NFT Businesses

• OpenSea’s historical peak of about $200 million in monthly fees was cited as an example of how rapid growth can be difficult to assess for durability.
• The speakers compared that pattern with excitement around AI businesses, arguing that impressive short-term growth does not necessarily mean a business has lasting economics.

Takeaways

• The discussion’s broader lesson is to examine whether growth is durable and supported by viable unit economics, rather than extrapolating from a short burst of activity.
• OpenSea was used as an example, not presented as a current investment recommendation.