Gartner Takes Another Stab At Forecasting AI Spending

Gartner extended its AI spending forecast through 2027 (dropping 2024 from view) and finds AI is already a major share of IT budgets — 31.7% in 2025 and 41.5% in 2026 — with a reasonable estimate implying AI could approach roughly 50% of IT spend in 2027. AI infrastructure and software spending are projected to nearly double over two years, though infrastructure growth is expected to slow in 2027 versus 2026, while spending on models, data-science/development tools, security, and AI data-management tools accelerates from smaller bases. The report provides less granularity (lumping client, edge and datacenter infrastructure together), which complicates precise allocation across CPUs, GPUs and XPUs for investment decisions.

Analysis

Market structure is bifurcating: Gartner’s view (AI = 31.7% of IT spend in 2025 → 41.5% in 2026 → ~50% in 2027) implies concentrated winners—accelerator vendors (NVDA, AMD), hyperscalers (AMZN, MSFT, GOOG) and specialist AI data/software (SNOW, MDB) will see outsized revenue growth and pricing power for compute capacity. Incumbent low-margin, non‑AI IT suppliers and firms whose products only “contain AI functions” will face budget reallocation and margin pressure as customers prioritize infrastructure and model/tooling spend.

Tail risks include rapid regulatory curbs on model monetization, a global chip supply shock, or a macro capex freeze; any of these could erase projected doubling of AI infrastructure by 2027. Near-term (days–weeks) risk centers on earnings guidance from NVDA/AMZN/MSFT; short-term (months) on supply-chain and capex orders; long-term (years) on energy/density limits and second-order effects like telecom bandwidth and data governance costs.

Actionable trade implications: bias long concentrated compute (NVDA, AMD) and cloud infra (AMZN, MSFT) while hedging with cybersecurity (CRWD, PANW) and data-management (SNOW) exposure; consider relative shorts in laggards to AI compute (INTC, select PC OEMs). Use pairs and option structures to manage timing: NVDA call spreads vs short INTC equity, or buy-monthly puts around hyperscaler earnings to guard against guidance misses.

Consensus blind spots: Gartner’s category inflation (counting modest AI features as “AI spend”) will overstate monetizable revenue for many software names — expect eventual re-rating for companies with <20% of revenue truly AI‑driven. Also, power/real‑estate constraints in datacenters and rising energy costs are underpriced risks that could cap gross margins for heavy compute providers.

Key Decisions for Investors

• Establish a 3% portfolio long in NVDA (ticker NVDA) via a 6‑month bull call spread (buy 1 ATM, sell 1.5x OTM) to capture continued GPU pricing power while capping premium; add another 1.5% if NVDA corrects >15% from current levels.
• Allocate 3–5% combined long to hyperscalers: 2% AMZN (cloud/C3 revenue exposure) and 2–3% MSFT (Azure + enterprise AI) via cash equity, hedge with 0.5% portfolio in monthly OTM puts around each company through next two earnings as event risk insurance.
• Establish a 1.5% long in Snowflake (SNOW) and 1% in CrowdStrike (CRWD) to play data-management and AI security tailwinds; trim or re-evaluate if Gartner or company-level KPIs show <20% revenue from AI within 12 months.
• Short 1–2% of portfolio in Intel (INTC) equity as a relative laggard in accelerator compute; alternatively run a pair: long NVDA (3%) / short INTC (1.5%) to express spread in compute share while capping directional beta.
• Monitor regulatory signals (EU/US AI rules, expected proposals in next 30–90 days) and energy/colocation metrics (wholesale power prices up >10% YoY or datacenter PUE deterioration) as triggers to reduce leveraged exposure by 50% within 5 trading days.