Almost half of data centers planned for 2026 in the US are likely to be delayed or canceled, and the reason is …

Nearly 50% of US data centres expected to come online by 2026 are likely to face delays or cancellations due to shortages of transformers, switchgear and batteries. Big Tech has committed over $650bn this year and projects consuming up to 12 GW are planned, yet only ~33% of projects are under construction, creating a widening gap between ambition and execution. Transformer lead times have stretched from 24–30 months pre-2020 to as long as five years, forcing import dependence (notably from China) and posing a material sectoral headwind that could slow the AI build-out.

Analysis

The immediate market implication is a durable shift in bargaining power and capex timing: hyperscalers will face a procurement premium and will triage projects by revenue density rather than chronological order, meaning ready-to-monetize AI workloads get priority while exploratory build-outs are deferred. Expect this to compress near-term incremental gross margins on new AI deployments and to slow the cadence of capacity-driven revenue inflection points that investors have been modeling into 2026–2027 numbers.

Second-order supply effects will reroute flows of capital and inventories into domestic assembly, refurbishment markets, and long lead suppliers; this creates a multi-year runway for onshore equipment OEMs to re-rate if they can convert backlog into throughput. However, execution risk is asymmetric: manufacturing scale-up requires 12–36 months of concentrated capex and stable input-costs, so equities here will likely trade on order-book visibility and policy milestones rather than on quarterly organic growth alone.

Geopolitical and policy catalysts matter more than usual — any credible Buy-American subsidy, tariff relaxation on specific subcomponents, or an S. legislative package that underwrites domestic transformers/energy storage would materially shorten timelines and reprice winners within 3–12 months. Conversely, export curbs or trade retaliation are persistent tail risks that could force hyperscalers to absorb higher operating costs or accelerate geographic diversification of AI workloads.

From an alpha perspective, the mispricing window is in the asymmetry between short-term pain and multi-year structural gain: equipment OEMs with credible ramp plans and backlog transparency are under-owned relative to the valuation hit across cloud infra names that will experience deferred monetization. Monitor order-book cadence, announced domestic capacity additions, and any swap of planned builds into colocation/outsourced capacity as binary catalysts.

Key Decisions for Investors

• Long GEV (GE Vernova) equity, 18–36 months horizon: size 3–5% of sector exposure. Rationale — exposure to onshore equipment buildout and repricing power as OEM order books tighten; target +35–60% if visible capacity ramps and large hyperscaler purchase agreements materialize. Downside: execution/commodity-cost slippage could compress upside by ~20–30%.
• Pair trade: Short AMZN (or buy AMZN 3–9 month puts) and Long MSFT (or buy MSFT 9–12 month calls) — net market neutral notional. Rationale — deferment of capex-driven AI rollouts disproportionately pressures AMZN’s retail-margin-levered valuation while MSFT’s software/SaaS cash flows provide ballast. Risk/reward: expect 5–15% relative outperformance for MSFT vs AMZN over 3–9 months if deployments are delayed; hedge with single-stock hedges sized 0.6x MSFT per 1x AMZN.
• Buy protective puts on META (6–9 months): 1–2% portfolio allocation. Rationale — higher sensitivity to scale economics in ad/AI monetization that could be pushed out by infrastructure delays; puts asymmetrically benefit if monetization lags. Cap losses limited to premium paid; upside if market reprices growth multiple down 10%+.
• Event-driven call spread on GEV tied to policy catalyst (12–18 months): buy 12–18 month GEV call spread (ITM to moderately OTM) funded by selling farther OTM calls. Rationale — concentrated payoff if a domestic subsidy/order-book acceleration is announced. Reward ~2–4x premium if catalyst hits; limited loss to net premium if it does not.