Episode

AI Bubble: Private Valuations vs Public Reality

An analysis of the divergence between AI startup valuations and public market multiples. The AI bubble is concentrated in private equity, where OpenAI and Anthropic trade at unsustainable levels compared to public peers like Nvidia. Upcoming IPOs will force a reckoning with true profitability.

The takeaways

AI Bubble Hides In Private Markets. The AI valuation bubble is not in public equities but in private startups like OpenAI and Anthropic. Public multiples have compressed, while private valuations remain detached from revenue realities.
Public Multiples Signal Cautious Optimism. Nvidia and Broadcom P/E ratios have fallen significantly, indicating investors are pricing in AI risks. This compression suggests the public market is not in a speculative frenzy.
Neocloud Debt Structures Are Unsustainable. Data center providers like CoreWeave face crisis-level yields due to borrowing at high rates for low returns. Rising infrastructure costs and slowing capital flows threaten their viability.
Adjusted Profitability Masks True Costs. AI labs report profitability by excluding training costs and revenue sharing. This accounting practice obscures the massive capital expenditure required to maintain product competitiveness.
IPOs Will Force Valuation Reckoning. Upcoming public listings will expose AI startups to rigorous public market scrutiny. Valuations may halve when compared against public peers like Nvidia, forcing price corrections or delays.
China Drives AI Price Deflation. Cheap open-weight models from China have cut token prices by half. This forces frontier labs to lower prices, squeezing margins while infrastructure costs rise.

The Divergence of AI Valuations

The current AI market presents a paradox: while headlines suggest a looming bubble, public market data indicates a cautious, risk-adjusted environment. The true speculative excess is confined to the private sector, where venture capital valuations for firms like OpenAI and Anthropic have decoupled from fundamental financial metrics. This divergence creates a significant risk for upcoming initial public offerings (IPOs), which will force these companies to confront the stricter valuation standards of Wall Street.

Public Market Compression

Contrary to popular belief, the AI bubble is not driving public stock prices to irrational heights. Major chipmakers such as Nvidia, Broadcom, and Micron have seen their price-to-earnings multiples contract by 48% to 56% over the past year. This compression reflects investor skepticism regarding the sustainability of AI-driven earnings, particularly given the high customer concentration and rising financing costs in the sector. While stock prices have risen due to exploding profits, the premium paid for those earnings has decreased, signaling a mature, risk-aware market rather than a speculative mania.

The Private Sector Bubble

The speculative frenzy is isolated in private markets, where OpenAI and Anthropic are valued at $1.4 trillion and $2 trillion, respectively. These valuations are supported by venture capital metrics that prioritize founder vision and talent over margins and profits. However, these companies are currently losing billions annually. When adjusted for true costs, including training expenses and revenue sharing, their profitability is questionable. Applying public market multiples, such as Nvidia’s price-to-sales ratio, would halve these valuations, revealing a significant overvaluation in the private sector.

Operational and Financial Risks

The infrastructure supporting AI, particularly neoclouds like CoreWeave, faces severe financial stress. High debt loads, combined with rising construction costs and slowing capital inflows, have pushed credit spreads to crisis levels. Furthermore, the entry of cheap Chinese open-weight models has deflated AI token prices, forcing frontier labs to reduce prices by up to 80%. This dynamic of rising costs and falling prices threatens the unit economics of the entire AI ecosystem.

Strategic Implications

Investors should distinguish between public AI stocks, which are priced with caution, and private AI startups, which are priced on hype. The upcoming IPO of Anthropic will be a critical test of this divergence. If management delays the listing or reduces the price, it will confirm the presence of a private market bubble. For entrepreneurs and investors, the focus must shift from narrative-driven valuations to sustainable unit economics and true profitability.

Key insights

The AI bubble is concentrated in private markets, not public equities. Public multiples for major AI players have compressed, reflecting risk awareness, while private valuations remain inflated.

Impact: Investors must adjust their risk models to account for the divergence between private and public AI valuations, particularly ahead of major IPOs.

Market Structure →

Neoclouds face unsustainable debt structures, with borrowing costs exceeding returns. This financial fragility threatens the infrastructure backbone of the AI industry.

Impact: Credit crises in data center providers could disrupt AI deployment and force a reassessment of infrastructure investment strategies.

Financial Risk →

AI labs use adjusted profitability metrics that exclude critical costs like training and revenue sharing. This obscures the true financial burden of maintaining competitive models.

Impact: Misleading financial reporting may delay necessary operational corrections and mislead investors about the sustainability of AI business models.

Accounting Practices →

Chinese open-weight models have deflated AI token prices, forcing frontier labs to cut prices by up to 80%. This squeezes margins while infrastructure costs rise.

Impact: Price competition from China threatens the profitability of US-based AI labs, potentially slowing innovation and capital investment.

Competitive Dynamics →

Upcoming IPOs will force AI startups to face public market scrutiny, likely resulting in significant valuation corrections. The gap between private hype and public reality is unsustainable.

Impact: IPO delays or price cuts will signal a broader correction in the AI sector, affecting investor sentiment and capital allocation.

Capital Markets →

Action items

Re-evaluate exposure to private AI startups by applying public market multiples to their valuations. Identify companies where private valuations significantly exceed public peer benchmarks.

Impact: This helps mitigate the risk of overvaluation in private holdings and prepares for potential write-downs during IPOs.

Monitor credit spreads and debt structures of neocloud providers like CoreWeave. Assess the impact of rising interest rates and construction costs on their viability.

Impact: Early detection of financial distress in infrastructure providers can help avoid indirect exposure to AI sector failures.

Scrutinize AI company financial reports for adjusted profitability metrics. Request detailed breakdowns of training costs and revenue sharing to understand true margins.

Impact: Accurate assessment of unit economics is crucial for making informed investment decisions in the AI sector.

Track the pricing trends of Chinese open-weight models and their impact on US-based AI labs. Adjust forecasts for revenue and margins accordingly.

Impact: Understanding competitive pressure from China helps in predicting the sustainability of AI pricing and profitability.

Prepare for potential IPO delays or price cuts by Anthropic and OpenAI. Develop contingency plans for investments tied to these companies' public listings.

Impact: Proactive planning for valuation corrections can protect portfolios from sudden market adjustments.

Quotes

“That's like taking out a margin loan to buy treasuries, Gil told us.”

“The bubble's not in the stock market. It is hiding somewhere else.”

“It's equivalent to me inserting lifts in my shoes and saying, I'm six foot six, but on an adjusted basis.”