Goldman Sachs Forecasts AI Capex Surge to Continue Through First Half of 2028
Summary
- • Goldman Sachs: AI supply-demand imbalance won't balance until H1 2028, driving sustained capex surge
- • Alphabet Q2 capex $44.9B; full-year guidance raised to $195–$205B with further increase expected in 2027
- • SpaceX Q2 capex $18.4B vastly exceeded ~$6B analyst estimates; JPMorgan models $200B/year AI spend
- • Memory prices, chip costs, and data center land acquisition all rising under AI supply chain constraints
Details
Goldman Sachs: AI supply-demand imbalance persists through H1 2028 — capex surge is multi-year
Eric Sheridan, Goldman Sachs TMT Group business unit leader, said the imbalance means continued capex, revenue growth, and supply chain constraints across the AI infrastructure stack for years.
Alphabet Q2 capex $44.9B, slightly above $44.7B Wall Street forecast; full-year guidance raised to $195–$205B
Full-year guidance raised from $180–$190B, with executives indicating a 'significant' further capex increase planned for 2027 as AI buildout continues.
SpaceX Q2 capex $18.4B — more than 3x analyst estimates of ~$6B; full-year ~$65B vs $50B Wall Street model
SpaceX's dramatic capex beat surprised Wall Street. JPMorgan separately warned SpaceX could allocate $200B to AI spending in each of the next two years.
Tesla committing $25B in capex for 2026, roughly 3x historical spend; further 2027 increase expected
Tesla's significant capex ramp is driven by Optimus humanoid robot and robotaxi production scaling — representing a new class of physical AI infrastructure investment.
Memory pricing, chip costs, and data center land acquisition all rising under supply constraints
Goldman Sachs noted memory pricing is increasing as the food chain remains constrained; chip prices are higher than 6–12 months ago; land for data center shells is being acquired aggressively ahead of future component availability.
Some Big Tech stocks, including Tesla, sold off during earnings season on spending concerns
Investors are increasingly questioning whether aggressive AI capex spending is already priced in, with some stocks falling on earnings results despite positive AI demand signals — indicating market skepticism about return timelines.
Source: Yahoo Finance / Goldman Sachs Research (Eric Sheridan interview), August 31, 2026.
What This Means
Goldman Sachs's forecast that AI infrastructure spending will remain elevated through early 2028 is a significant signal for the entire AI supply chain — from chipmakers and memory manufacturers to data center operators and energy providers. The persistence of supply-demand imbalance means pricing power on components like GPUs, HBM memory, and custom silicon will remain high, benefiting hardware companies while sustaining elevated costs for AI builders. For investors, the key question is whether this extended multi-year capex cycle will translate into revenue and earnings growth that justifies the spend — a question markets are beginning to ask more skeptically as some Big Tech stocks sold off on earnings spending announcements. The scale of SpaceX's capex surprise ($18.4B vs. $6B estimated) illustrates how consensus forecasts are still systematically underestimating AI infrastructure ambitions.