Morgan Stanley: AI infrastructure boosts its contribution to U.S. growth, while political and energy constraints loom larger

How much has AI investment’s contribution to U.S. GDP growth risen?

Morgan Stanley estimates that AI-related investment’s contribution to GDP growth rose from 0.2 percentage points in 2024 to 0.5 percentage points in 2025, reaching 0.9 percentage points so far in 2026. In the third quarter of 2026 alone, AI-related spending under the broader definition contributed 0.9 percentage points at an annualized rate, while the narrower AI-only measure contributed 0.6 percentage points.
More than three-quarters of related spending in the third quarter went to infrastructure rather than AI applications. The current boost to growth therefore comes mainly from construction and equipment investment; imports of related equipment mean that some of the growth effect of this spending occurs outside the United States.

As of the second quarter of 2026, output per worker in the United States was up 2.4% year over year, about twice the average growth rate of the past 20 years. Over the past four quarters, AI-exposed industries contributed 2.0 percentage points of that increase; a year earlier, these industries contributed 1.4 percentage points, while industries with low or medium exposure contributed a combined 1.0 percentage points.
The firm believes the productivity acceleration still mainly reflects capital deepening and output growth resulting from AI infrastructure investment, rather than widespread transformation of corporate processes. Some industries, such as data processing and software, have already reduced employment, but the firm estimates that AI has raised the unemployment rate by at most 15 basis points; labor substitution remains largely localized for now.

Morgan Stanley expects Democrats to gain greater influence at the state and local levels after the midterm elections, potentially bringing greater scrutiny of data-center approvals, energy costs and environmental standards. Possible measures include suspending construction, expanding local veto powers and tightening permitting requirements.
The firm still maintains its baseline estimate of about 1.6 trillion dollars in capital expenditure by hyperscalers in 2027, but believes political scrutiny is more likely to change project locations and timing than to end the investment cycle. Chip financing is relatively less affected, while new data centers are more likely to face approval delays; power, community opposition and supply-chain issues could also lengthen the time it takes for planned spending to feed through to GDP.

Morgan Stanley’s U.S. baseline forecast is GDP growth of 2.6% in 2026 and 2.4% in 2027 (both fourth-quarter year over year). It also expects AI-related investment to contribute about 0.6 percentage points a year in each of the two years. The firm also expects the Federal Reserve to raise rates by 25 basis points in December and again in March of the following year, lifting the target range to 4.25%–4.5%.
The oil-price shock has persisted for seven months, and buffer inventories are limited: as of 30 October, U.S. commercial crude inventories stood at 409.8 million barrels, down 1 million barrels in a week; the Strategic Petroleum Reserve held about 282.2 million barrels, its lowest level since April 1984. However, the SPR fell by only about 7 million barrels in October, less than the 30 million barrels in September, while U.S. crude production has recovered from earlier disruptions and refinery utilization has risen. Inventory pressures and a recovery in supply are occurring at the same time, providing key context for assessing whether oil-price shocks will continue to push inflation higher and, in turn, affect the growth and interest-rate outlook.