The AI Arms Race Is Shifting Again—These 2 “Picks and Shovels” Plays are Worth Adding to the Radar

$1 trillion in Vera Rubin and Blackwell orders through 2027 was highlighted by Nvidia CEO Jensen Huang, yet NVDA shares rose only ~1% on the day, suggesting muted market reaction. The piece flags Broadcom as a complementary 'picks and shovels' AI play despite a high 63.4x trailing P/E and suggests it could enter the top-three market caps within five years. ASML remains monopolistic in EUV lithography, unveiling a new machine that could boost production by ~50% by 2030, supporting continued demand for advanced silicon capacity; overall tone is bullish on AI infrastructure but cautious given valuation digestion.

Analysis

The immediate market pause after another bullish industry narrative is a digestion of concentrated capital and execution risk rather than a collapse in underlying demand; compute budgets are front-loaded into a multi-year buildout but procurement cadence will be uneven as customers sequence training vs inference investments. That sequencing benefits firms that sell purpose-built inference accelerators, advanced-node process tools, and interconnect/IP that scale across multiple customers — demand becomes broader across the stack even as unit growth for a single product family oscillates.

A key second-order effect is supply-chain bifurcation: constrained, specialized machine supply and advanced-node scarcity force fabs and system integrators to prioritize high-margin, high-volume customers — this amplifies pricing power for sole-source providers but raises concentration risk (order volatility, regulatory diversion). At the same time, increased custom silicon adoption lifts adjacent markets — HBM, advanced packaging, optical I/O and firmware/IP providers — creating durable multi-vendor TAM expansion rather than a single-winner outcome.

Tail risks live in three buckets and different horizons: near-term (0–6 months) earnings/stock-structure noise and inventory digestion; medium (6–24 months) execution on ramping new fabs and converting backlog into revenue; long (2–5 years) structural disintermediation if large cloud players push in-house silicon aggressively or if export/regulatory fragmentation curbs addressable markets. Positioning that captures the broadened infrastructure demand while hedging concentration and regulatory downside offers asymmetric payoffs over 12–36 months.

Key Decisions for Investors

• Accumulate AVGO on a 12–18 month horizon (target +25–35% upside, max downside -12–15% if hyperscaler in-sourcing accelerates). Entry: scale in on a 5–10% pullback or earnings-related weakness; size 2–4% NAV. Rationale: exposure to custom silicon/IP and networking that benefits inference deployments; use a trailing stop at -12%.
• Build a 24–36 month core position in ASML on dips (target +40–60% from current levels, downside protected by monopoly tech but subject to regulatory/export risk of -20–25%). Keep position small-to-moderate (3–5% NAV) and add on any 10% drawdowns tied to macro headlines rather than order shifts.
• Relative-value pair (6–12 months): Long AVGO + ASML (equal weights) / Short NVDA tactical trim. Finance by selling short-dated NVDA call spreads to offset cost. Objective: capture multi-vendor infrastructure re-rating while reducing single-name concentration; expect 1.5–2x asymmetric payoff if NVDA underperforms through inventory digestion.
• Options hedge for event risk (next 90 days): Buy NVDA 6–9 month puts or buy a protective put calendar (longer-dated put, short near-dated put) sized to cover delta exposure of core positions. Cost is insurance; breakeven justified if single-name degradation >15% during inventory reset.