Dimon Says AI Capital Spending Will Hit $725 Billion in 2026. Here Are the Sectors That Will Win and the Ones That Will Be Left Behind

Jamie Dimon projects the five hyperscalers will raise AI-driven capital spending from $450B in 2025 to $725B in 2026 — a ~61% increase — driving strong demand for data-center upgrades, chips, foundry capacity, and optical equipment. Winners include data-center REITs (Equinix, Digital Realty — yields ~1.9% and 2.7%), chipmakers (Nvidia, Broadcom), TSMC, and suppliers like Lumentum and Corning; losers are likely older/smaller cloud software vendors (Salesforce, ServiceNow, C3.ai) vulnerable to hyperscaler first‑party AI offerings. Portfolio implication: overweight AI infrastructure and supply‑chain plays, underweight legacy cloud/software exposed to displacement.

Analysis

The capex cycle is morphing from a pure compute-topping story into a systems problem: power, cooling, interconnects, packaging and long-lead fab capacity are the real bottlenecks that determine winners beyond raw GPU vendors. That favors vertically integrated suppliers with scale and fixed-capex leverage (foundries, ASIC houses, optical and heavy electrical equipment) because each incremental AI rack drives outsized demand for their adjacent products and services for multiple years.

On the software side, commoditization and bundling risk is underappreciated. Enterprises will increasingly accept second‑party AI primitives from large infrastructure providers in exchange for integration and price, compressing ASPs for specialist SaaS. Smaller AI software vendors that lack exclusive data, deep vertical workflows, or sticky billing mechanics face rapid margin erosion as hyperscalers push first‑party alternatives into procurement processes.

Key risks and timing: in the near term (weeks–months) macro volatility and inventory swings can create price dislocations in chips and REITs; medium term (6–18 months) architectural shifts toward lower‑power ASIC inference or on‑prem custom silicon would re‑rate GPU beneficiaries; long term (2–5 years) fab capacity growth and hyperscaler verticalization could erode third‑party leasing and component pricing power. Watch ASPs, utilization rates, hyperscaler leasing vs build disclosures, and new product cadence from ASIC/foundry partners.

A contrarian read: consensus overweights a single‑vendor GPU scarcity narrative and underweights the acceleration of ASIC adoption plus hyperscaler internalization. That creates asymmetric opportunities to pair concentrated hardware exposure with shorts in commoditized software franchises that lack defensible data moats.

Key Decisions for Investors

• Long NVDA via Jan-2028 LEAP calls (size 1–2% NAV): directional play on GPU-driven training demand with 18–36 month horizon. Trim or hedge if NVDA options IV compresses >30% or revenue guidance misses; target 2.5x upside vs 1x downside if stopped at 30% drawdown.
• Long TSM and AVGO (total position 3% NAV) to capture foundry and ASIC leverage; scale in on pullbacks >8% over 30 days. Hedge with a small NVDA call sale if you need to fund entry; hold 12–24 months and expect 40–80% IRR if secular demand persists.
• Pair trade: long EQIX (1.5% NAV) / short CRM (1.5% NAV). EQIX for sticky leasing cashflows and pricing power; short CRM to express risk of SaaS bundling. Rebalance at quarterly rent growth prints and CRM subscription/margin beats; target 1.8:1 R/R over 12 months.
• Tactical short: buy 6–9 month puts on C3.ai (size 0.5–1% NAV) as a hedge against software commoditization. Exit or convert to collars if M&A rumors or a major partnership surfaces. Risk: binary event upside—limit position to small hedge allocation.