Tech companies may only get half the profit they need to justify AI investment, Goldman analyst warns

The S&P 500 hit a record close at 6,944.82 (+0.62%) as markets price continued AI-driven capex from hyperscalers; Goldman Sachs forecasts a 12% total return for the S&P 500 to 7,600 by year-end 2026 but warns AI capex growth will decelerate and force rotations among mega-cap tech stocks. Goldman estimates hyperscaler capex was roughly $400bn in 2025 (up ~70% y/y) and says maintaining investor-expected returns would require an annual profit run-rate exceeding $1tn versus a 2026 consensus income of $450bn, creating two-way risk; policy tailwinds (OBBA full expensing), potential Fed easing and easier bank lending support more capex but increase leverage and profit requirements. Market snapshot showed mixed global indices and Bitcoin at $91.8k, underscoring upside driven by AI spending but material downside if profits fail to keep pace.

Analysis

Market structure: AI hyperscaler capex ($400bn in 2025, +70% YoY; consensus ~$500bn pa 2025–27) disproportionately benefits hardware (GPUs, servers), datacenter specialists and select cloud vendors while squeezing low-ROIC software and peripheral services. Concentration risk is rising: top 10 S&P names = 41% market cap and drove >50% of 2025 returns, so marginal capex deceleration will cause idiosyncratic re-pricings and greater dispersion within mega-cap tech over 6–18 months. Cross-asset: larger corporate debt issuance to fund capex points to heavier IG supply and tighter credit spreads if demand holds, while equity implied vol on tech should rise on rotation; energy and copper demand for datacenters creates modest upward pressure on power/commodity prices over 12–36 months.

Risk assessment: Tail risks include a) profit shortfall vs required >$1T run‑rate (Goldman math) causing >20% re-rating of overvalued names, b) regulatory limits on AI data usage or export controls disrupting revenue, and c) a credit shock if corporates can’t refinance incremental debt. Immediate (days) risk is profit-taking at record highs; short-term (weeks–months) is rotation-driven volatility; long-term (quarters–years) is structural winner-take-most outcomes and potential mean reversion in returns on invested capital. Hidden dependencies: OBBA tax treatment and Fed easing can prolong capex; energy/PCI constraints and GPU supply bottlenecks can cap marginal returns.

Trade implications: Favor concentrated hardware/AI infra longs (NVDA exposure via 3–6M call spreads) and selective longs in durable monetizers (GOOGL, AMZN, META) while funding these with shorts of high-valuation, low-ROIC software/data plays (e.g., SNOW-sized names) in a 6–12M horizon. Use S&P protective put spreads (3M, ~5–8% OTM) to guard against a >7–10% drawdown; consider selling covered calls on top-10 mega-cap names after 5–10% pop to harvest volatility. Size trades modestly (1.5–3% ticket risk each) given concentration/tail risk.

Contrarian angle: Consensus expects capex deceleration; that misses two scenarios — OBBA-driven extension of full expensing that could sustain above-consensus capex into 2027, or the opposite: diminishing returns forcing a rapid profit shock that forces a multi-quarter derating. Historical parallel: 2000’s tech capex boom-to-bust shows hardware suppliers can re-rate violently when end-demand falters. The market may underprice outcomes: either stronger hardware winners (buy NVDA) or rapid dispersion and debt-driven downside (protective hedges and selective shorts).

Key Decisions for Investors

• Establish a 2–3% portfolio position in NVIDIA (NVDA) exposure via 3–6 month call spreads (e.g., buy 1–2 strikes 5–15% OTM) to capture further AI compute demand while limiting premium outlay; trim on a 30% move higher or if GPU supply/demand signals weaken in next 60 days.
• Add a 1.5–2% overweight in Alphabet (GOOGL) and 1% in Amazon (AMZN) equities for 6–12 months as durable monetizers of AI, but sell covered calls (30–60 day) once position gains >10% to monetize elevated implied vol and lock-in returns.
• Implement a pair trade: long NVDA (1.5–2%) / short high-valuation data/cloud growth name(s) (e.g., SNOW or equivalent, 1.0–1.5%) with a 6–12 month horizon—target relative performance capture if capex growth decelerates and multiples compress.
• Purchase S&P 500 3-month put spread sized to 1.5–2% portfolio risk (example strikes ~5–8% OTM from current 6,944) to protect against a >7–10% market correction tied to capex-profit disappointment or credit shock; close if S&P >7,600 (Goldman YE forecast) or if consensus profit revisions improve >10% over 60 days.
• Set explicit monitoring triggers: if reported hyperscaler capex growth falls below +20% YoY or consensus corporate profit revisions for top-10 tech fall >10% over 60 days, reduce aggregate tech exposure by 30–50% and shift into cash/IG credit opportunities.