The tech stocks free fall doesn’t make any sense, BofA says in rebuke to investors while doubling down on the sector’s longevity

Bank of America Global Research argues the recent software-led selloff that erased roughly $300 billion of market value is internally inconsistent, contending AI capex cannot both collapse and simultaneously render legacy SaaS obsolete. BofA forecasts AI capex will quadruple to $1.2 trillion by 2030, views leading chip names (Nvidia, Broadcom, AMD, Credo) as trading near or below 1x PEG versus 1.5–2x for large-cap peers, and highlights physical supply constraints (power, land, data-center shells, advanced memory and optics) as a governor on overbuild risk, framing the pullback as a potential buying opportunity rather than evidence of secular decline.

Analysis

Market structure is bifurcating: winners are AI-infrastructure suppliers (NVDA, AVGO, AMD, CRDO, memory and optics suppliers) who gain pricing power from constrained compute, power and data‑center capacity; losers are high-valuation SaaS incumbents (CRM, NOW) whose pricing power and renewal dynamics face uncertainty. The market is effectively repricing time-to-monetization: infrastructure capex remains supply-constrained and should support ASP/margin expansion over 12–36 months, while SaaS revenues face near-term churn and contract repricing risk over the next 2–6 quarters.

Tail risks include an AI-capex collapse (>30% yoy cut) driven by disappointing model economics, tighter US/China export restrictions, or a major fab/power outage; these could compress semiconductor revenue by 20–40% in a stress scenario. Immediate impact (days) will be volatility and option skew; short-term (weeks–months) could see earnings-guide resets; long-term (2026–2030) the BofA $1.2T capex thesis remains plausible but dependent on power/land/fiber buildouts and memory supply evolution.

Trading implications: favor asymmetric, concentrated longs in infrastructure vs tactical, small-cap shorts in SaaS. Implement risk-defined option exposure (6‑month call spreads on NVDA/AVGO) and relative-value pairs (long NVDA/short CRM) to capture dispersion while limiting tail exposure. Rotate portfolio +200–300bps into semis/hardware, funded by -200–300bps from SaaS/long-duration names, scaling entries over 4–8 weeks and trimming into a 20–30% rally.

Contrarian view: consensus understates physical bottlenecks that create multi-quarter pricing power for leading silicon and optics — this implies current selloff is overdone for best-in-class fabricators but underdone for smaller SaaS firms with real AI-driven churn risk. Historical parallel: Jan 2025 DeepSeek reaction recovered within 6–9 months as buildout reaccelerated; conversely, a technology-monetization delay could punish crowded infrastructure longs >30%. Set hard thresholds: exit/hedge if infrastructure order cuts exceed 20% sequentially or if SaaS ARR downgrades exceed 10%.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA and a 2% long in AVGO, scale in over 4–8 weeks; use 6‑month call spreads 20–25% OTM to express upside while capping premium (target net cost <3% notional).
• Initiate a relative-value pair: long NVDA (1.0% notional) vs short CRM (0.75–1.0% notional) to capture infrastructure vs SaaS dispersion; exit if CRM outperforms NVDA by >15% in 60 days or CRM reports >10% AI-driven ARR upside on next earnings.
• Reduce aggregate SaaS exposure (CRM, NOW) by 200–300bps and redeploy into semis/hardware; implement 1–2% tactical short positions in CRM/NOW funded by profits from chip longs and maintain stop-loss at 25% adverse move.
• Buy a 6‑month 15–25% OTM call spread on AMD (size 0.5–1% notional) and a smaller speculative 0.5–1% long in CRDO equity for upside optionality; cap downside with defined-risk options and close or roll if orders from hyperscalers fall >20% quarter-over-quarter.
• Monitor three catalysts over next 30–90 days before re-allocating further: hyperscaler capex guides (Dell/HP/MSFT/AMZN), US export-control announcements on advanced nodes/optics, and next two quarters of SaaS renewal/ARR data; increase exposure only if hyperscaler guidance stays flat or positive and order flow does not drop >15%.