Goldman Sachs Projects $700 Billion in Artificial Intelligence (AI) Capex This Year. Here's My Top Stock to Buy.

Goldman Sachs projects $500B is likely and up to $700B could be spent in 2026 to build AI infrastructure, implying a sizable and concentrated demand surge for power. Brookfield Renewable has a ~13.5 GW demand pipeline with Microsoft and Google, broad global generation and storage assets, and long-term contracts that align with persistent AI-driven electricity needs. The company offers two share classes yielding ~3.9% (corporate) and ~4.9% (partnership), has delivered ~5% annualized dividend growth over the past decade, and targets 5–9% annual dividend growth going forward. This positions Brookfield Renewable as a durable dividend/renewables play to benefit from AI-driven power demand, likely influencing sector and stock-level performance.

Analysis

AI-driven data-center expansion will create concentrated, long-lived demand for grid capacity and firming — not just raw MW. The economic choke points over the next 12–36 months are likely to be transmission interconnection queues, permitting lead times, and the availability of dispatchable low-carbon capacity (hydro, nuclear, long-duration storage). Owners that can deliver contracted, location-specific capacity will capture outsized spreads versus intermittent generators selling into volatile merchant markets.

Second-order winners include companies with existing site control, water rights or vintage hydro assets, and firms that can monetize incremental services (synthetic inertia, capacity tags, dark-fiber collocation, behind-the-meter storage). Expect upward pressure on local capacity market clearing prices in constrained ISO zones, which will re-rate cash-flow-stable, contracted renewable platforms while compressing effective realized returns for unhedged merchant assets. Conversely, suppliers of commodity solar/wind with long interconnection delays or high curtailment risk are exposed to lower-than-forecast utilization.

Key tail risks: (1) an AI capex miss from macro tightening or a semiconductor-led efficiency shock that materially reduces power/compute intensity; (2) regulatory/political pushback that chokes transmission rollouts, concentrating demand but also increasing development costs; (3) a rapid, unexpected drop in battery LCOE that shifts economics away from hydro to cheaper storage faster than models assume. Near-term catalysts to watch are large cloud PPA announcements, FERC transmission rulings, and quarterly disclosures of contracted MW by hyperscalers.

There is a tradable structural arbitrage driven by investor constraints and tax/ticketing differences across share classes. That mispricing plus the profile of multi-technology asset owners creates asymmetric return opportunities if you size for development/time-to-contract execution risk.

Key Decisions for Investors

• Long BEPC units (12–36 month hold): allocate a modest position (2–4% NAV) to capture an asymmetric payoff from contract rollouts and re-rating of contracted cash flows. Hedge execution risk by trimming into large PPA announcements; worst-case scenario: 20–30% drawdown if PPAs slow or power prices collapse.
• Pair trade — long NVDA (buy 6–12 month call spread) / short INTC (buy 6–12 month puts or lean short): NVDA remains the primary beneficiary of any sustained AI capex cadence, while legacy fabs face margin/leverage risk. Target 2.5–3x reward-to-risk; close if NVDA total-return underperforms semis index by >15% in 3 months.
• Buy GS (6–18 month horizon): modest long to capture fees and advisory upside if large-scale capital raises and M&A accompany the buildout; size small relative to macro beta. Cut exposure on a macro shock that contracts bank capital markets activity (indicator: primary equity issuance down >40% year-over-year).
• Event-driven overlay: buy long-dated MSFT or GOOGL LEAP calls (18–36 months) against a small hedge of short-term out-of-the-money calls to fund cost — directional play on cloud/colocation demand growth. Exit or delta-hedge if cloud CAPEX guidance is reduced by >10% on next two quarterly calls.