Why I'm Buying Growth Stocks While Everyone Else Is Panic-Selling Tech

Hyperscalers are set to spend roughly $700 billion on AI infrastructure this year, supporting continued demand despite investor rotation out of tech following the war with Iran. Evidence cited includes 7-8 year-old Alphabet TPUs and 5-year-old CoreWeave GPUs remaining 100% utilized, rental rates down ~70% from peak but still monetized, and TSMC materially increasing capex to expand advanced-chip capacity. The author argues these signals validate continued AI-driven growth and is buying tech growth names (Alphabet, Amazon, Broadcom, AMD). Disclosures note positions in AMD, Alphabet, Amazon, Broadcom and The Motley Fool's related recommendations.

Analysis

The AI compute wave is creating layered, time-staggered demand: near-term pressure on high-end GPUs and networking, medium-term growth in custom accelerators and CPUs for agentic workloads, and multi-year structural upside for foundries and systems integrators that capture design-to-deploy lock‑in. That favors firms that sell platform-level solutions (chip IP + software + integration) and penalizes commodity suppliers whose product can be rapidly out‑optimized.

A subtle second‑order is an aftermarket capacity market: as hyperscalers internalize training, a bifurcated rental/spot market for older‑node silicon and specialized interconnects will persist, creating recurring revenue opportunities for niche cloud providers and colo partners while compressing spot pricing volatility. Another is capital intensity risk concentrating systemic exposure in a few foundries; supply shocks or financing stress there transmit far faster to the compute stack than to end‑user software demand.

Key risk triggers and timeframes: within 0–6 months watch hyperscaler utilization and spot rack leasing spreads for a meaningful demand inflection; 6–24 months watch foundry guidance, backlog-to-ship ratios, and margin migration from training to inference. A macro downturn that reduces enterprise AI budgets or a rapid shift to ultra‑low‑power, vertically integrated accelerators would materially reset the winners list and valuations.

Key Decisions for Investors

• Pair trade (6–18 months): Long AVGO (buy 6–12 month calls or 3–5% NAV outright) / Short INTC (buy 6–12 month puts). Rationale: capture networking/custom‑chip wins vs foundry/node execution risk. Target 25–40% gross return on the pair; cut if spread moves against you by 12% (stop).
• Long foundry exposure (12–36 months): Buy TSM shares or buy deep‑ITM 18–24 month calls sized to 3–6% NAV. Thesis: asymmetric upside from pricing power if capacity tight; hedge tail Taiwan/geopolitical risk with 12–18 month put protection (cost ~3–5% NAV).
• Tactical convexity trade (3–9 months): Buy AMD 9–12 month call spreads (delta ~0.35–0.6) to capture server CPU cadence without paying full premium. Risk management: sell into 30–50% move; max loss = premium paid.
• Contrarian idea (6–12 months): Accumulate GOOGL on pullbacks vs small‑cap value basket — pair to neutralize macro beta. Target asymmetric 20–35% upside with stop if cloud rev share/revenue growth decelerates two consecutive quarters.