A Generational Investment Opportunity: 3 AI Stocks I'm Buying Now

Nebius reported core AI revenue growth of 802% YoY in Q4 and projects an annual run rate of $7–9 billion by end-2026 (up from $1.25 billion at end-2025) while expanding data center sites from 2 in 2024 to 7 in 2025 and targeting 16 by end-2026. Broadcom and Nvidia are positioned to capture a large share of an estimated $3–7 trillion cumulative data-center capex opportunity through 2030; both are portrayed as underpriced on forward P/E with the market largely pricing only a successful 2026. The piece recommends buying AVGO, NVDA, and NBIS and notes Nebius shares are >20% off their all-time high, creating a buying opportunity.

Analysis

Broadcom and Nvidia are playing complementary roles in a multi-trillion-dollar capex cycle; the non-obvious effect is that hyperscaler co-designs (Broadcom-style) convert one-off chip sales into multi-year, high-visibility revenue streams that stiffen barriers for pure-play fabless vendors. That stickiness also crystallizes upstream demand for HBM, advanced substrates and interconnects — creating a durable pull-through for memory and packaging suppliers even if hyperscalers diversify accelerator stacks.

Nebius functions as an accelerant to that pull-through: a well-funded neocloud buying at scale smooths out demand volatility for accelerators and creates a private procurement channel that can pre-empt spot markets. The second-order consequence is tighter utilization of GPU/ASIC capacity and longer effective lead times for smaller customers, which compresses pricing elasticity and supports OEM margins in the near term but raises concentration risk if a handful of customers pause builds.

Key risks are asymmetric: model-level efficiency gains or a rapid shift to cheaper inference hardware could truncate demand within 6–18 months, while antitrust scrutiny of preferential supply deals could disrupt channel dynamics over 12–36 months. Near-term signals to watch are HBM pricing and backlog days (leading indicator), Nebius ARR composition and gross margin trend, and hyperscaler cadence calls on design-win fulfillment.

Contrarian take: the market is likely underestimating Nebius’s ability to convert preferential hardware access into durable ARR, but overestimating the timeline — expect meaningful binary outcomes across 12–24 months rather than a smooth linear ramp. For Broadcom, value is asymmetric if you’ve conviction in multi-year hyperscaler design-ins; for Nvidia, premiums are justified but vulnerable to short-term demand normalization or substitute accelerators.

Key Decisions for Investors

• Long NBIS (size 1–2% NAV): stagger entry 50% now, 50% on >15% pullback; horizon 12–24 months. Hedge downside by buying 9–12 month puts to cap loss to ~25% of position cost. Rationale: optionality on ARR conversion with capped downside.
• Buy AVGO via cash-secured puts (6–9 month expiries) sized to acquire 1–1.5% NAV at strikes ~5–10% below current market to collect premium and set a conservative entry. Hold 12–24 months; trim if design-win revenue misses guidance. R/R: collect premium now, upside if multi-year contracts re-rate.
• Directional NVDA exposure through 3–6 month call spreads (long ATM call, short OTM call) to capture momentum while limiting premium bleed; size small (0.5–1% NAV) and monitor HBM/backlog data. Overlay with a 12-month hedge: short a small INTC position (equal dollar) to protect vs semiconductor cyclic/architecture sideways risk.
• Pair trade: long AVGO / short INTC (equal dollar, 1% NAV each), horizon 12 months. Expresses structural shift to custom accelerators over CPU-centric stacks; cut if INTC reports accelerating data-center wins or Broadcom design-win cadence slows. Target relative outperformance 2:1 if secular trend continues.