There isn’t just one AI bubble, strategist says — there’s a ‘rolling sequence of bubbles’ instead

Dhaval Joshi argues the “AI bubble” is not one implosion but a rolling sequence of sector-specific booms and busts as investors reprice narratives. He cites near-tripling silver prices driven by data-center power demand optimism, plus a semis rebound on assumed chip pricing power that he expects to unwind as margins normalize. Key risks highlighted are a tightening in highly accommodative monetary policy (via sharp rises in real interest rates/real yields), a sharp capex unwind, or a non-mild recession—especially as hyperscaler capex is increasingly overtaking free cash flow (e.g., Google turning FCF negative for the first time).

Analysis

This is a dispersion signal, not a blanket AI-short. The first-order losers are the names whose valuations require both sustained top-line growth and continued capital intensity: MSFT, GOOGL, META, AMZN, and especially ORCL if the market starts discounting weaker FCF conversion or slower buybacks. The second-order effect is factor-level: if hyperscalers’ buyback capacity slows, a hidden source of index support for QQQ/XLK fades, which can pressure multiple expansion even before earnings revisions turn down.

The key catalyst path is rates plus capex cadence, not “AI sentiment.” A 10-20 bp rise in real yields, or any signal that 2026 capex stays elevated while monetization lags, should trigger a fast de-rating in the highest-duration tech names over the next 1-3 months. By contrast, a soft patch in growth would be worse for BAC/GS in absolute terms, so financials are a relative hedge only if the economy stays intact and rates stay firm.

The contrarian miss is that the market may be overpricing the idea that every dollar of AI spend is waste. A few platforms do own the tollbooths, and the durable winners may simply be the firms with enough distribution to pass through the cost. Falsify the bearish case if hyperscaler guidance shows capex flattening, cloud/ads monetization re-accelerates, or real yields roll over enough to re-expand long-duration multiples.

Key Decisions for Investors

• Sell 1-3 month QQQ or XLK call spreads on any post-earnings tech bounce; this is a tactical hedge against margin-compression risk, not a core short.
• Pair trade: short ORCL / long GS or BAC into the next earnings window; the spread expresses AI-capex disappointment versus a less capex-intensive financials hedge. Cover if ORCL shows improving FCF conversion or capex growth moderates.
• Trim META and AMZN before the next guidance cycle if management reaffirms aggressive spend; their valuation is most exposed to buyback/FCF slowdown over the next 1-2 quarters.
• Prefer MSFT and GOOGL over ORCL in any long-only AI basket; they have the best shot at capturing AI value through existing distribution rather than pure spend.
• Set a macro alert: if real yields break materially higher, add to tech hedges; if yields fall back sharply, cover shorts because the duration trade will likely reflate fast.