Exchange 2026: ALPS Talks AI Infrastructure Beyond the Chip

Three-fund "mosaic" approach: Paul Baiocchi (SS&C ALPS Advisors) recommends targeting AI infrastructure beyond semiconductors, focusing on the physical systems that power the digital/AI revolution. He advocates using three separate funds to capture distinct layers of infrastructure growth (compute, data-center/physical infrastructure, and enabling hardware/supply-chain). This is directional investment guidance rather than market-moving news; consider modest portfolio reallocation toward targeted tech and infrastructure exposures rather than broad semiconductor bets.

Analysis

The real money in AI infrastructure is being allocated to physical flow — electrons, cooling, fiber and heavy electrical kit — not just compute bricks. Expect outsized revenue growth for firms that control long lead-time, high-margin bottlenecks: large power transformers, high-voltage switchgear, industrial chillers and specialty copper/fiber suppliers; these components act as gatekeepers because lead times and permitting create a multi-quarter queue that front-runs chip-level demand.

Second-order beneficiaries include construction OEMs (cranes, modular-build specialists), firms that provide contracted long-term maintenance and SLAs, and utilities that can monetize interconnection queues via rate-base projects. Losers are the lowest-margin server OEMs and spot GPU resellers if hyperscalers internalize system integration and move to bespoke racks and DPUs; that re-shapes margins down the stack for commodity hardware vendors within 6-24 months.

Key risks are macro-driven capex compression, utility rate shocks and export/regulatory shocks that could stall multi-year builds; these can flip a 24-month growth story into a 6-12 month slowdown. Near-term catalysts to watch: hyperscaler capex guidance, multi-TSF land/lease transactions, transformer order backlogs and utility interconnection queue clearance rates — any of which will move the trade within days-to-weeks, while permitting and build completions play out over 12-36 months.

Consensus is overweighting chips and GPUs while underweighting the balance-sheet assets that actually limit deployment pace. That makes equipment and grid-exposure names asymmetric: they are underpriced for the optionality of multi-quarter capacity scarcity but vulnerable to rapid demand normalization if model count growth slows or financing costs spike.

Key Decisions for Investors

• Long Eaton (ETN) — buy 12-24 month call spreads or 2% of portfolio in equity: thesis is ETN captures increased demand for power distribution and transformers; target +30% if utility/grid capex accelerates in 12-24 months, downside -15% if capex freezes. Enter on any 5-10% pullback or after a hyperscaler capex acceleration print.
• Long Freeport-McMoRan (FCX) — accumulate over 3-9 months via cash-secured puts or small outright long: copper is a choke-point for cabling and grounding; expect 6-18 month re-rating if data-center siting continues, skewed reward ~3:1 vs downside that correlates with broad risk-off.
• Long Digital Realty (DLR) or Equinix (EQIX) selectively — prefer names with long-term contracted revenue: buy on dips or after an announced large land/lease deal; thesis captures pricing power from scarce interconnected space. Risk: REIT multiples already price growth, so cap position size to 1-2% and hedge with short cyclicals if macro tightens.
• Pair trade: long modular data-center integrators / short commodity server OEMs — implement via long A (AMT/ETN exposure) and short H (commodity OEM ETF or select names) over 6-18 months to capture margin reallocation from servers to integration and site services; set stop-loss at 12% adverse move and take profits if differential widens >20%.