Debunking The Bulls' Main Arguments On AI

The article argues semiconductor stocks, including Nvidia, are in a dotcom-like bubble, with AI profits overstated by circular capital flows between hyperscalers and startups. It warns that cyclical demand and circular cloud spending commitments are masking underlying weakness and implying unsustainable valuations. The message is broadly negative for AI-linked equities and sentiment across the semiconductor sector.

Analysis

The key second-order issue is not just that semis are expensive; it’s that a meaningful share of the current AI capex cycle is being financed through a closed loop that can temporarily suppress downside in suppliers while worsening the eventual adjustment. If startup funding and cloud commitments are the real collateral supporting AI optimism, then reported demand is front-loaded and brittle: the market is capitalizing several years of spend as if it were recurring cash flow. That setup typically ends with multiple compression first, then a slower air-pocket in orders as the weakest participants fail to raise the next round.

For NVDA, the risk is less about next quarter and more about the 6-18 month path of estimate revisions once hyperscalers start optimizing utilization instead of adding capacity. The beneficiaries of a slowdown are the platform owners with balance-sheet flexibility and pricing power, while the most exposed names are the second-tier GPU ecosystem, networking, packaging, and venture-backed AI compute buyers that depend on perpetual financing. A break in circular funding would also hit private markets harder than public markets initially, because late-stage marks are more sensitive to forward revenue assumptions than to current revenue.

The catalyst set is asymmetric: a modest miss, reduced capex guidance, or one high-profile startup down round can trigger a reassessment of AI TAM far faster than the underlying spending actually rolls over. The contrarian risk is that the market may already know this and still be underestimating how long hyperscalers can keep spending from operating cash flow; in that case, a short too early gets punished by another 1-2 quarters of blowout orders. The cleanest signal to watch is whether cloud vendors begin emphasizing utilization, payback periods, and internal AI monetization rather than raw deployment volume.

This is a classic “funding structure” trade, not just a valuation trade: when financing conditions tighten, the narrative can unwind faster than the fundamentals. If the circularity thesis is right, the first visible cracks will be in venture funding, then in smaller model labs, then in supplier order books; by the time margin pressure shows up in semis, the trade will already have moved.

Key Decisions for Investors

• Short NVDA on strength over the next 1-3 weeks, using call spreads or a small outright short; target a 10-15% drawdown if the market starts discounting 2026 capex normalization, with a hard stop above the prior momentum high.
• Pair trade: long cash-rich hyperscaler with the strongest free-cash-flow conversion / short NVDA or a basket of AI hardware suppliers for a 3-6 month horizon; thesis is that owners of demand will outperform suppliers as capex efficiency improves.
• Buy 6-12 month downside protection on NVDA via put spreads rather than naked puts; the implied move is still vulnerable to a one-quarter squeeze, but convexity becomes attractive if guidance commentary shifts toward utilization discipline.
• Avoid chasing venture-linked AI beneficiaries and private-market proxies for the next 1-2 quarters; funding round risk is highest there, and marks can reprice 20-40% before public-market semis fully react.
• Set an alert for any major hyperscaler commentary on AI ROI, utilization, or capex pacing; a pivot from 'build' to 'optimize' is the highest-probability catalyst for a multi-week de-rating in the whole AI complex.