AdvertiseFee Deal Desk

Microsoft

Goldman Sachs Sees Hyperscaler AI Capex Peak at 1.5% of GDP in Q4 2026

Goldman Sachs sees hyperscaler AI investment growth peaking near 1.5% of GDP in Q4 2026, with Amazon, Microsoft and Alphabet capex at $725B to $800B.

Be a creator
November 20, 2026, 08:33 PM UTC4 min read
AI SummaryAI
  • Goldman Sachs projects hyperscaler investment growth to peak near 1.5% of GDP in Q4 2026.
  • Goldman Sachs estimates annual US AI capital expenditure at about $600 billion.
  • Amazon, Microsoft and Alphabet 2026 capex projected between $725 billion and $800 billion.
  • Goldman estimates AI spending added roughly 0.1 percentage point to US GDP growth.

Goldman Sachs Puts AI Capex at $600 Billion

Goldman Sachs expects the increase in investment by US hyperscale cloud providers to peak at roughly 1.5% of gross domestic product in the fourth quarter of 2026, then fade, even if the absolute level of AI spending stays elevated. The bank's research puts total US AI-related capital expenditure at about $600 billion a year, a figure that covers chips and hardware as well as the physical facilities built to house them. The peak window, in the bank's reading, stretches from the end of 2026 into early 2027.

Amazon (AMZN), Microsoft (MSFT) and Alphabet (GOOGL) anchor the spending side. Goldman's range for the three hyperscalers' 2026 capital expenditure runs from $725 billion to $800 billion. Hyperscale cloud providers, in the bank's usage, are the companies that sell large-scale computing services and expand the facilities behind them directly.

Chief economist Jan Hatzius has been skeptical about how much of that spending actually accrues to the US economy. His argument rests on import content: when a US company buys chips and hardware made abroad, the outlay counts as domestic investment, but the import value is deducted from GDP, shrinking the net effect. Goldman Sachs estimates AI spending has added about 0.1 percentage point to US GDP growth, and in some periods the contribution has been close to zero. Public estimates that credit AI capital spending with 30% to 90% of US growth sit well above the bank's arithmetic. Goldman treats the total investment figure and the GDP contribution as separate indicators, and the new forecast turns on what happens to the contribution when the growth rate of spending decelerates.

The 1.5% Is an Increment, Not a Level

The 1.5% figure needs careful reading. It is not total AI investment measured against the size of the economy; it is the ratio of the year-on-year increase in hyperscaler investment to GDP, which is why the bank reads it as a peak-and-fade path rather than a plateau. Growth responds to how fast spending rises, not to how large spending has become. The same scale of investment, sustained at a constant level, keeps production high but stops adding anything new to the growth rate.

Import dependence adds a second discount. AI infrastructure leans heavily on chips and hardware produced overseas, so corporate outlay and domestic value added do not match, a structure that runs directly through the way AI server imports are subtracted in GDP accounting. Company revenue, domestic production and macro GDP can therefore diverge even when the spending headlines look identical.

The bank frames AI infrastructure construction as a structural change spread over decades, not a one-off investment burst, so a slower ramp does not mean adoption stops. Productivity improvement, in this reading, arrives as firms apply AI in day-to-day operations, and the investment growth rate and the productivity payoff need not move at the same time. What to track from here is concrete: the quarterly capital expenditure guidance Amazon, Microsoft and Alphabet publish, and the US GDP statistics. Whether corporate spending tracks the 2026 range of $725 billion to $800 billion, and what the growth contribution does after the fourth-quarter peak, are the markers the bank names for the next phase. The same arithmetic implies readers weighing each capex announcement should look at the increase and the import share together, not the headline amount alone.

AMZN Slides Toward Strongest Support

Markets are already trading the deceleration argument on Friday: COINOTAG data shows Amazon at $247.11, down 4.37% over the past 24 hours and sitting just above its strongest support level at $246.95, which scores 99/100 on COINOTAG's composite. Microsoft is at $485.18, down 3.15%, with the next support at $475.25 scoring 85/100. Perpetual futures tied to Amazon carry a funding rate of 0.0007%, and open interest in Microsoft perpetuals stands at $27.8 million. A daily close below $246.95 would put Amazon's deeper levels in play; holding it keeps the stock range-bound while the capex debate runs.

COINOTAG's editorial and research desk.

AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.