This Is The Problem With AI's Most Bullish Chart
This Is The Problem With AI's Most Bullish Chart
1 hour ago•Odd Lots•Bloomberg
Podcast1 hr 9 min
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • Treat NVIDIA (NVDA) as a watch-list investment, not a straightforward AI-growth bet: monitor whether neocloud demand is supported by customer revenue or continued financing, and whether NVIDIA scales back support.
  • Approach AI neoclouds cautiously; tighter lending, lower loan-to-value ratios, and customer prepayments point to financing and execution risks, while CoreWeave (CRWV)’s customer commitments and financing access are useful indicators—not proof of safety.
  • If considering Anthropic after its expected public debut, look past IPO-demand headlines and assess whether enterprise revenue can cover compute costs and infrastructure spending.
  • Palo Alto Networks (PANW) and CrowdStrike (CRWD) are potential beneficiaries of rising software vulnerabilities, but the discussion offers no specific price targets or buy recommendations.
Detailed Analysis

NVIDIA (NVDA)

  • The discussion portrays NVIDIA as central to AI infrastructure financing as well as chip sales: the company has invested in neocloud equity raises and provided support to creditors.
  • A guest described this financing as a cycle in which support for compute providers can help them buy more NVIDIA equipment, while venture capital funds some of the demand for compute.
  • The hosts and guests raised concerns that this cycle could unwind quickly if financing tightens. NVIDIA’s management was described as potentially becoming more cautious about supporting neoclouds, which could add stress to the market.
  • The discussion also mentioned a ransomware group’s claim that it had accessed NVIDIA customer and shipment data; the transcript did not establish whether that claim was verified.

Takeaways

  • The transcript offers a cautious, risk-focused view of NVIDIA’s role in the AI boom, rather than a specific stock recommendation.
  • Investors assessing the AI trade could watch for changes in NVIDIA’s support for neoclouds and for evidence that compute demand is being funded by customers’ revenue, rather than mainly by venture capital.
  • The reported cyber incident is an additional concern raised in the discussion, but its verification was not established in the transcript.

CoreWeave (CRWV)

  • CoreWeave was presented as an example of a neocloud responding to tighter financing conditions by requiring larger customer prepayments and seeking loans backed by higher-credit-quality customers.
  • The guest said public markets rewarded these steps. He contrasted CoreWeave’s financing position with smaller neoclouds that may have less experienced operators and greater reliance on customer prepayments.
  • The broader neocloud market was described as facing lower loan-to-value limits, higher borrowing costs, and increased scrutiny of operators’ ability to meet service commitments.

Takeaways

  • CoreWeave’s ability to secure financing and customer commitments may be an important indicator of conditions in the wider AI infrastructure market.
  • The discussion highlights financing and execution risks for neoclouds; it does not establish that CoreWeave is insulated from those risks or make a specific recommendation on its shares.

Anthropic

  • Anthropic was expected to begin trading publicly the day after the podcast was recorded, though the hosts noted that the timing was not certain.
  • The bullish case discussed was that Anthropic could sell compute-intensive AI services at positive unit margins, particularly to enterprise customers.
  • A guest argued that Anthropic’s business may depend increasingly on vertical products in areas such as legal, finance, and healthcare, rather than only training frontier models.
  • The discussion also raised several concerns: Anthropic may need to spend heavily on data centers, may compete with companies that provide its distribution, and may face challenges converting AI usage into sustainable revenue.
  • One guest cautioned that IPO “oversubscription” figures are non-binding expressions of interest and, in his view, do not by themselves demonstrate durable investor demand.

Takeaways

  • The transcript presents Anthropic as a high-interest but uncertain AI investment, with enterprise revenue and the economics of compute as key questions.
  • For a prospective investor, the discussion suggests looking beyond IPO demand headlines and examining whether revenue can support compute costs and future infrastructure commitments.
  • No price target or specific buy recommendation was given.

AI Neoclouds and Compute Financing

  • The episode’s main investment theme was that AI infrastructure is also a credit and financing story. Neoclouds borrow to buy GPUs, while some customers rely on venture funding to pay for compute.
  • Loans described in the discussion included examples with loan-to-value ratios that had fallen from about 70%–80% toward 50%–60%, alongside higher interest costs and customer prepayments.
  • The guests warned that GPU rental rates may reflect not only underlying demand but also the cost and availability of financing. Spot-market prices were described as a thin and potentially vulnerable reference point.
  • Potential risks discussed included defaults, difficulty transferring compute contracts, inexperienced operators, customer dependence on venture funding, and a sudden increase in equipment being re-let after failures.
  • The guests also discussed possible financial infrastructure for the sector, including transferable compute contracts, resale markets, and put-option hedges tied to GPU rental indices. These were described as developing market mechanisms, not as established protections.

Takeaways

  • Treat rising GPU rental rates as an incomplete signal: they may reflect financing constraints as well as strong end-user demand.
  • The discussion suggests monitoring whether AI companies can pay for compute from operating revenue, rather than relying mainly on equity funding.
  • Private-credit exposure to neoclouds appears particularly sensitive to operator experience, customer credit quality, loan terms, and the legal ability to re-let capacity after a default.

Cybersecurity — Palo Alto Networks (PANW) and CrowdStrike (CRWD)

  • A guest argued that AI tools can make it easier to identify software vulnerabilities, while also pointing to a sharp increase in reported critical software bugs.
  • He described this as a potential tailwind for cybersecurity companies, naming Palo Alto Networks and CrowdStrike.
  • The discussion also raised the risk that software outages and security problems could increase as AI-generated code and rapid bug fixes become more common.

Takeaways

  • The episode identifies cybersecurity as a potential beneficiary of AI-related security and software risks, but it does not provide company-specific financial analysis or a stock recommendation.
  • The opportunity is tied to the possibility that organizations will need more security tools as vulnerabilities and operational disruptions increase.

Reinsurance — Swiss Re and Munich Re

  • The guest said reinsurers, including Swiss Re and Munich Re, may benefit from uncertainty around AI-related risks because insurers may have difficulty assessing and pricing those risks.
  • He characterized the current environment as favorable for reinsurers’ pricing power, while noting that AI risks could be very large and difficult to quantify.

Takeaways

  • Reinsurance was presented as a possible indirect beneficiary of growing AI-related risk and uncertainty.
  • The discussion does not quantify the potential impact on these companies or recommend either stock.

Copper

  • Copper was mentioned as a commodity whose price had fallen about 5% that day. The speakers noted that copper’s chart had recently moved in a way that resembled the Nasdaq and semiconductor stocks.
  • Copper was used as an example of how a commodity linked to a popular investment theme can become a proxy for sentiment, rather than a clean measure of underlying fundamentals.

Takeaways

  • The discussion provides no direct copper recommendation. Its main implication is to be cautious about treating a commodity’s price movement as clear evidence of AI demand or economic value.

Microsoft (MSFT)

  • Anthropic was described as selling a significant share of its products through Microsoft Marketplace, while also moving into areas where it could compete with Microsoft’s own products.
  • The guest said that this creates a potential conflict: a distribution partner may be less willing to promote a supplier that is becoming a competitor.

Takeaways

  • The episode highlights a distribution and competition risk for technology platforms that partner with AI companies that may later enter adjacent product markets.
  • No stock-specific recommendation was made.

Sun Microsystems (Historical Analogy)

  • Sun Microsystems was cited as a historical example of a company that sold servers to internet-era startups. The comparison was used to illustrate how investment in a promising technology cycle can create strong demand for a particular category of equipment without guaranteeing that all buyers or suppliers will ultimately succeed.

Takeaways

  • The analogy is a reminder that equipment demand during a technology boom does not, on its own, establish that the financing behind that demand is sustainable.
  • Sun Microsystems was discussed as a historical comparison, not as a current investment opportunity.
Ask about this postAnswers are grounded in this post's content.
Episode Description
It has been a historic week for AI: the Anthropic IPO and the ongoing situation at Nvidia headline a pretty head-spinning time for the industry. Despite Nvidia's earnings outlook looking good, and despite the scale of the Anthropic IPO — it's oversubscribed, as you might know — there are so many open questions about the state of the industry right now. Financing the buildout has become increasingly complicated, and the Democrats who won races in the midterms know that data center opposition is a winning issue. Today's episode is a special two-parter that addresses some of these questions. First, we speak with Stanley Lee, the chief product officer at Liquid Compute, and he explains why Wall Street and Silicon Valley seem to be talking past each other, and what fragilities are emerging in the credit market for GPUs. All those charts that show rising GPU rental rates, Lee tells us, might not be as industry-bullish as they first appear. Then, we turn to Paul Kedrosky, a managing partner at SK Ventures, to discuss why “number goes up and to the right” does not necessarily prove the viability of the capex involved in the AI buildout, and also if Nvidia will remain the central bank for AI. Read more: Anthropic Said to Be Weighing IPO Price as High as $85 Per Share Nvidia Sizes Up Arm CEO Haas in Search for Huang Successor Only http://Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at  bloomberg.com/subscriptions/oddlots Subscribe to the Odd Lots Newsletter Join the conversation: discord.gg/oddlots See omnystudio.com/listener for privacy information.
About Odd Lots
Odd Lots

Odd Lots

By Bloomberg

<p>Bloomberg's Joe Weisenthal and Tracy Alloway explore the most interesting topics in finance, markets and economics. Join the conversation every Monday and Thursday.</p>