Wealth Manager: “Hard to See a World Where You're Not Investing in AI” Buy These Dips Now

NVIDIA reported Q4 FY2026 Data Center revenue of $62.3B (+75% YoY) and full-year FY2026 revenue of $215.94B (+65%), while Anthropic's ARR climbed to north of $30B from $9B at the end of 2025. Several mega-cap AI/infra names have pulled back YTD—Microsoft ~-23%, Meta ~-13%, Broadcom ~-9%, QQQ ~-4%—even as Meta, Microsoft and Alphabet guide 2026 capex of $115–135B, ~ $120B (annualized) and $175–185B respectively. Analyst consensus target for NVIDIA is $268 vs current $177.64; the article frames current discounts as a long-term buying opportunity in AI infrastructure despite flagged regulatory risks and elevated VIX-driven volatility.

Analysis

The buildout is shifting value downstream from pure semiconductor design to vertically integrated infra stacks and long-term hyperscaler commitments. Hyperscalers’ multi-year buildouts create durable demand for high-end GPUs, networking ASICs and memory, but they also concentrate bargaining power with a handful of suppliers — winners will be firms that combine proprietary silicon with software lock‑ins (higher margin capture), while standalone component suppliers face more volatile orderbooks and inventory cycles.

Key tail risks live on three horizons: near-term volatility around earnings and options expiries (days–weeks), medium-term re-pricing if capex slows or open-source models reduce cloud compute intensity (3–12 months), and structural policy or supply constraints from export controls or power/cooling limits (12–36 months). A single large hyperscaler pivot to an alternative architecture or an aggressive price war on inference services could wipe out projected capacity-led revenue growth across the chain within a 6–12 month window.

Consensus is underweight the margin impact of non-chip costs (power, networking, real estate) and overestimates linear scaling of ARR into vendor revenue — many deployments substitute model engineering for raw scale once a baseline infra exists. Positioning should therefore favor convex exposure to execution (option structures on leading silicon vendors) and relative-value trades that isolate compute demand resilience (search/ads/cloud franchises) from social/engagement‑driven ad cycles.

Key Decisions for Investors

• NVDA — Buy a 9‑12 month call spread (buy Jul/Aug 2026 1st OTM / sell 2nd OTM) sized at 3% portfolio notional. Rationale: captures continued data‑center momentum with defined max loss (premium) and 2–3x upside if server rollouts continue; cut half at 50% realized premium loss and take profits at 100–150% gain.
• GOOGL / META pair — Go long GOOGL shares (2% portfolio) funded by an equal‑dollar short on META (2% portfolio), horizon 6–12 months. Rationale: hedge index/tech beta while expressing conviction in diversified ads/search resilience; target 20–30% relative outperformance, stop the pair if the spread moves against by 10% or if either issues materially different FY capex guidance.
• MSFT — Buy and write (buy shares, sell 6–9 month +10–12% OTM calls) sized 2–3% portfolio. Rationale: generates carry while retaining upside to continued AI platform monetization; acceptable return 8–12% annualized, unwind on a >12% downside breach or if guidance misses materially.
• AVGO — Sell 12‑15 month 8–10% OTM cash‑secured puts (size 1–2% portfolio) to collect premium with intent to own at a discount. Rationale: captures yield while taking optionality to accumulate semiconductor networking exposure on pop; max assignment risk is purchase at strike minus collected premium, close positions if AVGO gap declines >20% or if supply‑chain commentary turns structurally negative.