US economic outlook and AI's effects on growth and consumption Report Interpretation

Overview

This US Economics Weekly examines how AI investment is affecting growth and consumer cohorts, alongside the latest readings on oil, financial conditions, tariffs and GDP tracking. Morgan Stanley argues that AI is already a significant growth driver, but that the transition from infrastructure build-out to adoption will determine both the durability and distribution of its economic effects.

Core views

Morgan Stanley refines its estimate of AI investment's contribution to US GDP by expanding the investment categories included, improving its import treatment, and distinguishing broad AI-related expenditure from spending incremental to the AI boom. This distinction matters because corporate capex can include imported and intermediate goods, while GDP measures domestically produced final output. Imported data-center equipment may be part of capex plans without directly adding to US GDP. After adjusting for re-exports and estimating the share of computer imports used for personal consumption rather than business investment, the report estimates that broad AI-related investment has added an average 0.6 percentage points to annualized real GDP growth since 2025; the narrower incremental AI-only measure has added about 0.4 points. The current AI impulse is still led by physical build-out. Of roughly 0.5 percentage points contributed by AI-only investment in 1H26, around 0.4 points came from infrastructure—including data centers, computing equipment, networking hardware, power infrastructure and related machinery—with the rest from adoption, for which software investment is the report's economy-wide proxy. Infrastructure's contribution accelerated from approximately 0.15 points in 2023 and -0.05 points in 2024 to 0.21 points in 2025 and 0.36 points in 1H26. Morgan Stanley expects the mix to change as the cycle matures: infrastructure creates capacity for AI applications, while software spending signals broader diffusion across businesses. Software contributed about 0.18 points to growth in 2026, versus 0.15 in 2025 and 0.21 in 2024. The report expects infrastructure growth to slow in 2027-28, while adoption spending proves more durable and potentially accelerates. The report argues that AI's consumer effects cannot be judged through occupational exposure alone. High-income, college-educated, city-dwelling households—described as CHIC households—are disproportionately represented in highly AI-exposed occupations, but exposure can result in either automation or productivity-enhancing augmentation. Unemployment has risen more in high-exposure occupations, with displacement so far more severe for younger workers. For experienced workers, however, AI may raise productivity and wages, while new AI-related jobs have been concentrated among similarly educated and higher-income consumers. Wealth provides an additional offset for affluent households. The top 20% income cohort, college-educated households and people aged 55 or older each hold roughly 80% of household equity wealth. For the highest-income cohort, equity wealth is approximately six times annual labor income, so relatively modest asset-price gains can counter the consumption effect of a temporary labor-income decline. Lower-income households have less equity wealth in dollar terms and relative to their overall wealth, making spending less responsive to equity-market movements. Price effects differ by horizon and consumer group. In the near term, AI build-out can raise demand for electricity, software and chip-intensive goods. Morgan Stanley does not expect these relatively small categories to overturn its near-term disinflation view, but notes that this AI-exposed basket is a larger share of spending for lower-income, less-educated and younger households. Over time, productivity gains should be disinflationary, particularly where adoption is quicker. Higher-income households devote relatively more spending to financial, professional and communications services and may therefore experience AI-related services disinflation earlier. In Morgan Stanley's base case, AI causes a modest and temporary increase in unemployment of around 0.2-0.5 percentage points at the peak, but task creation, wage gains and wealth effects outweigh displacement. CHIC households benefit most on net, although established older professionals may be better placed to capture productivity and asset-price gains than highly educated younger professionals. The key downside scenario is materially faster diffusion than workers and firms can absorb: displacement would initially hit highly exposed white-collar households, then weaken broader demand, making monetary and fiscal responses important. Outside AI, the weekly trackers point to an energy and financial-conditions headwind. US crude and petroleum-product inventories were broadly unchanged week over week as of September 4, while Strategic Petroleum Reserve crude inventories fell to around 285 million barrels, their lowest level since April 1984. At a recent decline pace of about 30 million barrels every four weeks, the report estimates inventories could approach the roughly 70 million-barrel conservative operational minimum by end-February 2027. Following renewed US-Iran escalation, spot prices as of September 8 were $94.21 per barrel for WTI and $106.12 for Brent. Domestic crude production has trended higher since early May, while crude exports have declined and net oil-product imports have moved up from their April low. Morgan Stanley's FRB/US-based financial-conditions measure indicated that conditions as of the September 10 close were tighter than before the latest Middle East escalation and no longer easier than before the July FOMC meeting. The index combines the 10-year Treasury yield, S&P 500 returns, BBB corporate credit spreads, the US dollar and oil prices using their estimated growth elasticities relative to the federal funds rate. Since hostilities began on February 28, the tightening was equivalent to about a 53-basis-point increase in the federal funds rate, driven mainly by higher Treasury yields and oil prices; buoyant equities partly offset the restraint. On tariffs, the report estimates the effective rate at about 6.5% in July and an average 6.7% across May-July, and expects it to rise toward roughly 10% by year-end. The final forced-labor Section 301 action, effective July 24, imposed a 10% duty on economies with adopted, partially adopted or committed forced-labor import prohibitions and 12.5% on most others, while retaining Section 232, USMCA-compliant and other exemptions. Morgan Stanley views the structure as consistent with its expectation that Section 301 will operate as a broad country-level tariff floor. It does not expect Section 338 tariffs applied to about $20 billion of Canadian imports to materially change its year-end estimate. The report also explains its use of CBP-related Daily Treasury Statement withdrawals as a high-frequency proxy for tariff refunds. Morgan Stanley's 3Q GDP tracking estimate fell one-tenth to 2.0% after existing home sales reduced its residential-investment estimate. This tracker is a mechanical aggregation of monthly activity data feeding into the BEA calculation, not the institution's official GDP forecast. The Atlanta Fed's estimate remained 4.7%, with its goods-consumption estimate of 2.3% about 2.25 points above Morgan Stanley's and its services estimate of 4.5% about 1.5 points above; it also assumes faster inventory investment. The New York Fed nowcast rose 0.1 point to 2.3%, mainly on strong August payrolls. Morgan Stanley expects the upcoming data calendar to clarify the Fed outlook, retail demand, housing and manufacturing, while emphasizing that inflation data will determine the Fed's next move.

Analysis framework

The report combines national-accounting adjustments for AI capex with a decomposition of infrastructure and software-adoption spending. It then assesses consumer transmission through labor, household wealth and prices, and supplements that framework with high-frequency oil, financial-conditions, tariff and GDP trackers. Its GDP tracker mechanically aggregates monthly data inputs to the BEA framework, while its financial-conditions index translates market moves into a federal-funds-rate-equivalent effect on activity.

Methodology notes

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  AI infrastructure versus AI adoption decomposition

  The report separates data-center, computing, networking, power and machinery investment from software-based adoption to show how AI spending reaches GDP at different stages of the build-out.

• Other

  GDP accounting adjustment for imports and incremental AI spending

  Morgan Stanley adjusts capex for imports, re-exports and consumption uses because corporate spending does not translate one-for-one into domestically produced GDP.

• Other

  FRB/US-based financial conditions index

  The index aggregates five market variables using estimated growth elasticities and expresses their combined economic effect as an equivalent change in the federal funds rate.

Key data

• Broad AI-related investment contribution to real GDP growth0.6pp average since 2025Annualized real GDP-growth contribution
• AI-only investment contribution to real GDP growth0.4pp average since 2025Narrower incremental spending measure
• AI-only contribution in 1H26roughly 0.5ppAbout 0.4pp came from infrastructure
• Peak AI-related unemployment increase in the base casearound 0.2-0.5ppTemporary increase
• SPR crude inventoryaround 285 million barrelsLowest level since April 1984
• WTI and Brent spot prices$94.21/bbl and $106.12/bblAs of September 8
• Financial-conditions tighteningabout 53bpFederal-funds-rate-equivalent tightening since February 28
• Effective tariff rate~6.5% in July; 6.7% average in May-JulyExpected to converge toward ~10% by year-end
• Morgan Stanley 3Q GDP tracking2.0%Down 0.1pp in the latest week

Impact & implications

The report presents AI as an increasingly important source of US growth, with the next phase depending on whether software adoption broadens beyond the infrastructure build-out. It expects the benefits to be concentrated initially among affluent and established workers through wages, new jobs, wealth and services disinflation, while younger and lower-income households face relatively greater exposure to displacement or near-term price pressure. Higher oil prices, tighter financial conditions and rising tariffs are countervailing macro forces.

Risks

• AI diffusion could move materially faster than workers and firms can adjust, causing displacement among exposed white-collar workers and eventually broadening the demand shock.
• Higher oil prices and renewed Middle East escalation could intensify inflation concerns and tighten financial conditions.
• The Strategic Petroleum Reserve could approach its operational minimum by end-February 2027 if inventories continue to fall at the recent pace.

What to watch

• The pace at which AI spending shifts from infrastructure toward software adoption.
• Labor-market outcomes for highly AI-exposed workers, particularly younger entrants.
• Oil inventories, oil prices and the effect on inflation and financial conditions.
• The effective tariff rate and tariff-refund flows.
• Upcoming CPI data, which the report says will be important for the Fed's next move.
• Retail sales, housing data and industrial production for updates to GDP tracking.