AI Infrastructure Spending Could Nearly Triple by 2029. Here Are 2 Stocks to Buy.

Statista projects AI infrastructure spending will rise to $902 billion by 2029 (from $334 billion in 2025), underpinning ongoing data-center demand. Nvidia-anchored infrastructure remains dominant: data-center revenue was >90% of sales and grew 75% YoY, with the company reporting $120 billion net income on $215 billion revenue last year and citing $1 trillion in cumulative Blackwell/Rubin purchase orders through 2027. Iren is positioned on the power bottleneck side with >4.5 GW secured power, a $9.7 billion Microsoft contract, and a $3.4 billion annualized revenue target by end-2026 (only ~10% of its secured capacity), offering substantial long-term upside. Key risks include cyclical data-center spending, project financing and potential equity dilution.

Analysis

Nvidia’s de facto platform position creates a two-tier market: vendors that sell into the GPU stack (networking, racks, software partners) will see sticky revenue but also face sudden demand shocks if hyperscalers pause procurement. The much less-discussed beneficiary is owners of grid capacity and delivery — long lead-time electrical infrastructure (transformers, substation builds, PPAs) is now a scarce input, giving vertically integrated data‑center builders pricing optionality and faster monetization than traditional colo players.

Execution and financing are the primary vectors that will separate winners from losers over the next 12–36 months. Deal timing matters: a single delayed permit or a capital‑market tightening can push multigigawatt rollouts past hyperscaler budget windows, compressing returns and forcing dilution. Conversely, firms that can guarantee turnkey power and timing will command >200–300bp pricing premiums on multi-year capacity contracts.

From a competitive standpoint, the real long-term threat to the Nvidia monopoly is not an immediate GPU competitor but architectural substitution: in-house hyperscaler ASICs, tighter co-design between AI models and custom silicon, and networking disaggregation. Those shifts take 2–5 years to materialize, creating a window where capital should favor asset owners with secured power and fast build capability rather than pure-play hardware vendors whose revenue is more cyclical.

Trade execution should therefore balance secular optionality against cycle risk: buy exposure to monetizable real assets (power + racks) while expressing hardware-concentration risk via time-limited derivatives on Nvidia. Position sizing must assume 30–50% drawdowns in market selloffs and plan for 12–36 month holding periods to capture large hyperscaler deployments.

Key Decisions for Investors

• Long IREN (IREN) — size 2–4% NAV, target 2.5x upside over 24–36 months if company converts secured power to contracted revenue; protect with 12-month 15% OTM puts sized to limit loss to 20%. Rationale: monetizable power optionality and speed to market command premium; key risk is financing/dilution.
• NVDA options calendar — buy 24–30 month LEAPS call spread (bull-call spread) funded by selling 3–6 month calls to trim premium. Aim for asymmetric payoff where breakeven requires ~20–30% appreciation over two years while short calls harvest theta in near-term cyclical windows. Risk: abrupt capex pullback can trigger >30% drawdown in short term.
• Pair trade for a soft-landing scenario — long IREN / short NVDA near-term gamma (buy IREN equity 2–3% NAV, offset with NVDA short-dated 10–20% OTM call overwrites). This captures infrastructure premium if hyperscalers prioritize delivery speed while limiting net exposure to GPU cycle reversals.
• Catalyst watch & exit triggers — set alerts for (1) hyperscaler procurement pauses or macro funding stress (treasury yields >4.5% persistently) and (2) any 25% quarter-over-quarter slowdown in NVDA order flow; trim 30–50% of positions on either trigger within days-to-weeks to manage cycle risk.