Is the AI Bubble Bursting or Is the Sector Just Catching Its Breath?

Despite a short-term December correction and headlines around data-center financing and vendor payments, the article argues the AI investment thesis remains intact as hyperscalers and chipmakers (notably Nvidia and Broadcom) generate real revenue and profits from AI workloads. Goldman Sachs expects 2026 AI capex to exceed current forecasts, and bottlenecks beyond chips—energy, materials, data centers and memory—create additional investment opportunities (Micron shares have more than tripled over the past year). The piece favors small- and mid-cap AI beneficiaries for long-term upside even as large-cap leaders continue to capture the bulk of headlines.

Analysis

Market structure: AI capex is concentrating economic rent in GPU/SoC makers (NVDA, AVGO) and cloud providers (MSFT, AMZN, GOOGL) while creating a secondary winners’ list in memory (MU), data-center REITs, and energy/commodity suppliers (power, copper, lithium). Short-term losers include legacy software vendors with opaque AI monetization (ORCL flagged by headlines) and small AI services firms facing financing squeezes; pricing power for premium GPUs and HBM memory should keep ASPs elevated for 6–18 months.

Risk assessment: Tail events include export controls on advanced nodes/AI chips, a high-profile AI safety/regulatory shock, or a credit freeze for hyperscaler data‑center projects—each could crater demand within 30–90 days. Hidden dependencies: grid capacity and HBM supply are bottlenecks that can delay deployments; watch power-usage intensity metrics and wafer fab utilization rates as early indicators of stress.

Trade implications: Tactical exposure favors concentrated, size-controlled longs in NVDA (core), AVGO, and MU with position sizes 1.5–3% of risk capital each and trailing stops of 10%; hedge with a 0.5–1% short in ORCL to express payment/contract risk. Use 3–6 month call spreads on NVDA to express near-term upside while selling 12–20% OTM calls to fund premium; add 12–24 month LEAPS on MU for asymmetric memory upside.

Contrarian angles: Consensus underweights energy and memory constraints and overweights headline GPU winners—expect rotational outperformance from small-cap infra and power names if GPU rollouts accelerate. Historical parallel: unlike dot‑com, revenue/profitability exists here, so corrections will be volatility events, not structural failures; conversely, rapid capex could push real rates and compress multiples, an underappreciated macro drawback.

Key Decisions for Investors

• Establish a 2% portfolio long in NVDA within the next 2–6 weeks, scaling in on any 5–15% pullback from recent highs; protect with a 10% trailing stop or buy a 3–6 month call spread (buy ATM, sell 12–15% OTM) sized to limit max premium to 0.5% portfolio risk.
• Initiate a 2% long position in Broadcom (AVGO) and a 2–3% long in Micron (MU) as 6–18 month core holdings to play memory/HBM and interconnect demand; set target exits at +50% and stop-loss at -20% for each.
• Open a 1% short position in Oracle (ORCL) as a hedge against delayed enterprise AI monetization and OpenAI/contract-payment headlines; cover on signs of confirmed multi-quarter SaaS/AI contract wins or after 6 months.
• Rotate 3–5% from legacy IT services into data-center and energy infra names (data-center REITs, power utilities serving hyperscalers) over 1–3 months; add if hyperscaler capex guidance rises >10% YoY in quarterly reports.
• Monitor three catalysts over the next 90 days—(1) NVDA/AVGO earnings and guidance, (2) Goldman/industry capex revisions indicating >10% upgrade, (3) any export-control announcements—and tighten risk limits (reduce gross exposure by 30%) if two catalysts flip negative.