PodcastEmpireJason Yanowitz

Is The AI Debt Bubble Finally Popping? | Weekly Roundup

The discussion weighs the scale and funding risks of AI investment against uncertain returns, including questions about Anthropic's cash flow, market share, and the durability of AI businesses. Speakers also consider how AI could benefit traditional enterprises, while warning that deployment and productivity gains may take time. In crypto, the conversation spans stablecoin and tokenized-equity activity, Bitcoin's security funding and displacement risks, and the possibility of sharp volatility in public-market AI stocks.

Artificial IntelligenceTechnology StocksMarket ValuationsRisk ManagementStablecoinsTokenization

Preview · 21 more points in the app

Read the Full Summary

Key Takeaways

Speakers describe unstable-currency users turning to foreign currencies as a “life raft,” with U.S. stablecoins drawing stronger demand than Brazilian stablecoins; most Brazilian users in Nubank Crypto reportedly flock to USDC, and stablecoins make up about 75% of the wallet’s volume. David Vélez separately predicts that the number of global currencies could fall from 190 to 10 over 10 to 20 years.

The AI buildout could push related CapEx from roughly $1 trillion to $4 trillion by 2030, and speakers warn that Meta, Amazon, and Google could exhaust free cash flow and need debt financing. That spending pressure coincides with reported market-share losses for OpenAI and Anthropic, while Anthropic’s reported Q3 free-cash-flow positivity was questioned because its language was adjusted and it continued spending on training and frontier models.

Anthropic’s enterprise APIs are described as its main revenue source, with SemiAnalysis estimating an 89% gross margin, while subscriptions are presented as a route to convert users into API customers. Those figures sit alongside a report that Anthropic spent $1.60 on compute per dollar of revenue in the prior year, and a speaker’s skepticism about its reported Q3 free-cash-flow positivity.

Speakers say traditional enterprises could gain margin and productivity benefits from cheaper intelligence, but stress that the timing is uncertain and recent model-performance gains are becoming more marginal despite continued large investment. A speaker also says AI deployment remains very limited but expects it to grow by orders of magnitude, while noting that open-source models such as DeepSeek can still carry high inference costs.

Speakers caution that rapid growth and useful products do not guarantee durable AI businesses, and one compares some of the investment rush to 2021 while explicitly distinguishing AI from NFTs. One speaker predicts tier-four AI companies could be wiped out nearly immediately when the market turns, tier-three companies could fail over the following year or two, and places good SaaS businesses that cannot pivot in tier two.

GEO-POLITICS

Currency Rivalry — A speaker says the U.S. and China are seeking to expand their influence, and expects people in countries with unstable currencies to turn to another country's currency as a “life raft.” David Vélez separately predicts the number of global currencies could fall from 190 to 10 over 10 to 20 years.

“He said, there's 190 currencies around the world. My view is that in 10 to 20 years, those 190 will be 10…”—

TECHNOLOGY

AI Funding Risks — Jensen is cited as expecting AI-related CapEx to rise from roughly $1 trillion to $4 trillion by 2030. Speakers warn that spending could exhaust the free cash flow of Meta, Amazon, and Google, pushing them to tap debt markets. One speaker expects painful restructurings if companies cannot keep raising capital.

“The market is trying to make sense of this technology. It's trying to make sense of like a trillion dollars of CapEx…”—

Anthropic Economics — Anthropic reportedly spent $1.60 on compute for every dollar of revenue in the prior year. A speaker questions its reported Q3 free-cash-flow positivity, noting that the language was adjusted while Anthropic continued spending on training and new frontier models.

“And I think what is also true, and this is sort of an aside, but, you know, if you looked at Anthropix, some of the, the elite financials from last…”—

AI Market Share — A speaker says OpenAI and Anthropic are losing market share to open-weight models, and that Anthropic is also losing share to OpenAI. Both companies still face pressure to train frontier models and spend on CapEx.

“Because if that's true, there's this race happening right now where OpenAI and Anthropic continue to be like, oh, we have to be out on the frontier…”—

MARKETS

Crypto Resilience — The speaker says Bitcoin and the broader crypto market held up relatively well as equity markets, particularly AI stocks, faced a sharp selloff and high volatility.

“incredibly volatile you know stuff's down 20 30 10 days are normal surprisingly bitcoins held up quite well and the rest of crypto has been holding up quite well but…”—

CoreWeave Credit — At one point that morning, CoreWeave's credit default swaps were said to imply a 60% chance of default over the next few years.

“The CDS, so the credit default swaps in CoreWeave, I think at some point this morning were implying that the market believed there was a 60% chance that CoreWeave defaulted…”—

Inference Costs — A speaker argues that open-source models still carry inference costs and says running models such as DeepSeek can be expensive.

“So even if you look at like deep seek and some of these other open source models, like the cost to run that inference is quite high, sometimes even harder.”—

FORECASTS

AI Stock Volatility — One speaker expects public-market AI names to remain more volatile than crypto until Anthropic's S-1s are released and the market understands its free-cash-flow status. The speaker warns of possible 10% moves in a single day and advises against leverage.

“My view is, look, the market is going to continue to be extremely jittery and concerned until we the S1s come out…”—

Intro

The hosts open with light banter before noting that AI-related equities have sold off sharply while Bitcoin and broader crypto have held up better than expected.

AI Sells Off, Crypto Holds Up

Speakers connect AI-stock volatility to doubts about Anthropic’s growth and market share, massive projected infrastructure spending, and uncertainty over whether returns can justify the capital committed.

Is CoreWeave Facing A Debt Crisis?

Rob describes rising borrowing costs and default concerns at CoreWeave, warning that AI infrastructure firms may struggle to fund both growth and debt obligations if major model customers slow down.

Can Frontier Models Make Money?

The discussion weighs whether Anthropic and OpenAI can earn durable profits through enterprise API sales, despite high compute costs, uncertain free cash flow, and competition from open-weight models.

Ads (Fidelity Crypto, DAS London)

Mid-roll advertisements promote Fidelity Crypto and the Blockworks Digital Asset Summit in London.

Which AI Businesses Survive?

The speakers distinguish likely long-term AI winners from vulnerable companies, arguing that businesses reliant on continual capital and compute spending may fail as funding conditions tighten.

Who Wins As AI Gets Cheaper?

They favor traditional enterprises that can use falling AI costs to improve productivity, while questioning whether frontier-model performance gains are valuable enough to support enormous investment.

Ads (Fidelity Crypto, DAS London)

Mid-roll advertisements again promote Fidelity Crypto and the Blockworks Digital Asset Summit.

What Stood Out At Token2049?

At Token2049, Rob describes a more institutional conference focused on financial rails, stablecoins, tokenization, prediction markets, Robinhood Chain, and interest in Pearl and Zcash.

Can AI Mining Challenge Bitcoin?

The panel discusses Bitcoin’s security budget as miners move toward AI workloads and explains Pearl’s effort to pair AI inference payments with blockchain security, while debating its incentives and token economics.

Does Pearl’s Mining Model Work?

Rob presents Pearl as a way to reward useful compute and potentially attract hash power, while the hosts question whether inference work and token incentives can reliably align.

What’s Next For Crypto Exchanges?

The speakers review Backpack, Hyperliquid, and exchange growth, debating whether U.S. regulation will prevent shared global liquidity pools and how offshore demand compares with U.S. competition.

Will Stablecoins Replace National Currencies?

They debate whether stablecoins could displace weaker national currencies, citing Nubank’s stablecoin adoption while noting that governments’ monetary sovereignty and ability to issue currency remain powerful constraints.

Content Of The Week

In their closing recommendations, the hosts discuss a book on Genghis Khan, a Nubank interview, Nathan Fielder’s new film, and an upcoming conversation with investor Michael Dempsey.

Related Episodes

Important Disclaimers

Not Investment Advice: For education only — not financial, tax or legal advice.

AI-Generated: Made by AI from source transcripts and may contain errors.

No Guarantees: Past performance does not predict future results; investing carries risk.

Our Commentary · Request Removal