The AI Bubble Is Popping: Why These Stocks Are Crashing

A Dec. 15, 2025 Motley Fool video explains recent pullbacks in AI-focused equities, naming Nebius Group (NASDAQ: NBIS), Iren and CoreWeave and using Dec. 15 trading prices to illustrate the moves. The presenter attributes declines to sector sentiment and positioning rather than announcing new fundamental data, and discloses personal holdings in the mentioned names while promoting Motley Fool’s Stock Advisor service, indicating potential bias to consider when assessing the narrative.

Analysis

Market structure: The recent pullback is a rotation out of high-beta, small/mid-cap AI names (NBIS, IREN, small CRWV exposure) toward mega-cap AI enablers (NVDA, cloud GPU providers). Direct winners are GPU/IP owners and hyperscalers that control supply (NVDA, select CRWV customers); direct losers are balance-sheet-light software plays and regionally exposed utilities/industrial names that priced AI optionality into valuation. The demand signal is bifurcating — sustained enterprise GPU demand supports incumbents while discretionary AI experiments are being cut, compressing multiples of speculative names.

Risk assessment: Tail risks include tightened export controls on AI chips, acute GPU supply shocks, or an enterprise capex pullback (each could move prices 20–40% in 1–3 months). Immediate risk (days) is flow-driven volatility and gamma squeezes; short-term (weeks–months) depends on Nvidia guidance, year-end cloud bookings and Q4 results; long-term (quarters–years) is secular AI adoption but concentrated on a few suppliers (single-vendor dependency is a hidden fragility). Watch Nvidia ASPs, cloud utilization rates and borrow costs for names like NBIS — those are 30–60 day leading indicators.

Trade implications: Favor capital-efficient exposure to enablers: consider 4–6 month NVDA directional via a buy‑call/ sell‑call spread (long 30-delta, short 10–15% higher) sized 1–3% portfolio. For speculative small caps (NBIS, IREN) use asymmetric hedges: size shorts to 0.5–1.5% notional or buy 3-month 25–10 put spreads to cap downside while collecting premium, and consider a pair trade long CRWV vs short NBIS to capture dispersion. Rotate 20–30% of small-cap AI weights into infrastructure (CRWV, NVDA) and sell short-term covered calls on winners to harvest IV.

Contrarian angles: Consensus is likely over-penalizing fundamentals-light names — a >30% selloff in NBIS/IREN in 2–4 weeks without news often precedes takeover interest or re-ratings; distressed M&A is a realistic upside catalyst. Historical parallels: 2016–18 AI run-ups where hardware winners consolidated share; unlike prior cycles, current supply constraints and enterprise budgets give incumbents more durable pricing power. Beware: crowded shorts can squeeze if one of these small names secures a large cloud contract or positive regulatory clarity.

Key Decisions for Investors

• Establish a 1–3% portfolio position in NVDA via a 4–6 month call spread: buy a 30-delta call and sell a call 10–15% above current price to cap cost; target 30–50% downside-adjusted return, trim at +25–35% or if NVDA falls >12% below the 50-day MA.
• Reduce aggregate small-cap AI exposure (NBIS, IREN) by 20–40% over the next 10 trading days; where downside protection is needed, buy 3-month put spreads on NBIS (long 25-delta put, short ~10% lower strike) sized to 0.5–1.5% notional to limit tail loss.
• Implement a pair trade: go long CRWV (1–2% position) and short NBIS (0.5–1% position) to play dispersion; add if NBIS rises toward prior resistance or CRWV drops >10% on headline-driven flow. Use stop-losses at 12–15% adverse move.
• Sell 30–45 day covered calls on existing large-cap AI positions (NVDA, NDAQ exposure) to monetize elevated near-term IV; avoid selling deep OTM calls if IV is below 20% and underlying is <5% from 50-day MA.
• Monitor three specific triggers in the next 30–60 days before adding risk: Nvidia quarterly guidance vs consensus (GPU ASPs), cloud utilization metrics from AWS/Google earnings, and any new export-control/regulatory announcements; if all three are benign, add to enablers; if any are adverse, increase hedges by 50%.