Steve Eisman gets more conservative with portfolio as AI trade nears 'tipping point'

Steve Eisman says the AI trade is nearing a “tipping point” but is not net short, having hedged positions to reduce portfolio risk. He cited a decline in Anthropic’s revenue-to-equipment-expense ratio from 2.08 in Q2 to 1.7 in Q3 and expects its revenue growth to have decelerated, while noting its initial IPO filing omitted Q3 revenue figures. Alphabet shares fell after earnings as AI queries pressure search-ad monetization; Google said AI Mode ad experiments show promising results.

Analysis

The key risk is not simply that frontier-model growth slows; it is that investor expectations and infrastructure spending have outrun the evidence of durable monetization. If revenue per unit of compute deteriorates, model providers may need to choose between lower prices, slower capacity build-outs, or more external financing—each of which can weaken the expected returns embedded in the AI supply chain. The reported Anthropic figures are second-hand and require validation against consistent definitions and periods; incomplete IPO disclosure is a transparency warning, not proof of a deceleration.

For Alphabet, AI answers create a monetization-design problem: replacing a high-intent search result with a less monetizable interface can dilute economics even as engagement rises. Successful ad formats could reverse that pressure, so the issue is conversion and revenue per query, not query growth alone. Spillovers could reach AI infrastructure suppliers if customers defer deployment, but this evidence does not yet establish a broad capex reversal.

Near term, expect headline-driven volatility and position reduction rather than a clean, isolated AI short: crowded technology exposure can transmit losses into broader equities. Over 1–3 months, Anthropic’s fuller disclosure and Alphabet’s evidence on AI-query monetization are the useful tests. Over 6–18 months, the structural question is whether inference costs fall faster than pricing and whether AI products create incremental paid demand. The contrarian point: a tipping point in expectations need not mean a collapse in adoption; lower-cost models could expand usage while shifting value away from frontier-model providers.

Key Decisions for Investors

• Avoid a naked short of the AI complex. For the next 1–3 months, consider a small, defined-risk relative-value position that underweights GOOG against a diversified equity benchmark; size it as a hedge, not a high-conviction short, given the risk that improved ad formats restore monetization.
• Treat ANT as a disclosure-driven watch, not a short solely on the cited analysis. Reassess when the filing provides comparable quarterly revenue, equipment expense, cash burn, and customer concentration; deterioration across those measures would strengthen the downside case.
• Monitor GOOG revenue per AI-mode query, paid-click trends, and management commentary on ad tests. Stabilization or improvement in monetization would falsify the near-term bearish thesis; rising engagement without better economics would reinforce it.
• Use any broad AI-supplier exposure reduction selectively. Evidence of customer capex delays or weaker infrastructure orders is needed before extending this thesis to semiconductor or power-related suppliers; otherwise the move risks confusing model-provider economics with end-market adoption.