Goldman Sachs forecast that annualized spending on AI-related hardware and intellectual property will climb to $1.4 trillion by the end of 2027, equivalent to nearly 4% of the US economy, Reuters reported on 12 November. The bank forecast that spending would exceed $900 billion by the end of 2026, and said: “AI will remain the dominant driver of business investment growth in 2027”.
By our arithmetic on Bureau of Economic Analysis figures, information processing equipment and software accounted for about 64% of the $534 billion rise in US business investment in the year to the third quarter of 2026.
Business Pill · Capex and Depreciation Explained
An evergreen Business Engineer explainer on capital expenditure and depreciation. It explains the general idea only and does not discuss the Goldman forecast.
The key insight: As we read it, Goldman’s forecast extends a shift the national accounts already record: by our arithmetic on BEA data, computers and software supplied about 64% of the past year’s rise in US business investment, and Goldman says AI will remain the dominant driver in 2027.
Goldman estimates that AI-related investment will directly add 5.3 percentage points to 2027 capex growth, including semiconductor investment, and 0.6 percentage points to US GDP growth, according to Reuters. Excluding semiconductors, the bank put the boost at 4.3 points for capex growth and 0.2 points for GDP growth.
Overall, Goldman forecasts US business investment growth of 8% in 2027, which Reuters described as on a quarterly basis and nearly double the consensus forecast of 4.2%. Goldman said Democrats’ victories in this month’s midterm elections pose only limited risks to AI investment, Reuters reported.
The forecast comes amid concerns that rising borrowing costs and investor scrutiny could curb AI-related capital spending as companies lean more on debt, Reuters wrote. A separate Goldman trading-desk note last week argued that tech giants may pull AI capex forward into 2027 before planning rules change.
Year-over-year rise in US business (nonresidential fixed) investment, in billions of dollars at an annual rate, current dollars. In the year to the third quarter of 2026 it rose $534.3 billion: information processing equipment $270.6 billion, software $68.7 billion, everything else $195.0 billion; in the year to the third quarter of 2025, $268.7 billion. Our arithmetic on BEA data (third quarter 2026 = advance estimate of 29 October), via FRED.
What the National Accounts Already Show
In the third quarter of 2026, US companies invested $933.9 billion at an annual rate in information processing equipment and $831.5 billion in software, according to BEA’s advance estimate of 29 October. Together that is about $1.77 trillion, or 35.7% of the $4.95 trillion of business investment, by our arithmetic.
That share is the highest in the quarterly series, which begins in 1959, by our arithmetic; a year earlier it was 32.3%, and before 2025 the high was 32.1%, in late 2021. The figures are in current dollars, from BEA data published on FRED.
Business investment rose by $534.3 billion between the third quarter of 2025 and the third quarter of 2026. Information processing equipment supplied $270.6 billion of that, up 40.8%, and software $68.7 billion, by our arithmetic: about 64% between them, and about half from equipment alone. Everything else rose by $195.0 billion, or 6.5%.
In the year before, the two categories supplied about 74% of a smaller $268.7 billion rise. After adjusting for prices, business investment was 8.4% higher in the third quarter than a year earlier, by our arithmetic on BEA’s chained-dollar series; the Reuters report does not say whether Goldman’s 8% is in real or nominal terms.
Information processing equipment plus software as a share of US business (nonresidential fixed) investment, current dollars, quarterly from 2019 to the third quarter of 2026. Our arithmetic on BEA data via FRED; the third-quarter 2026 share of 35.7%% is the highest since the series begins in 1959.
Why GDP Gets Less Than Capex
Goldman’s GDP figure is far smaller than its capex figure, and the bank gave the reason: “Semiconductor investment is undercounted in the business investment statistics, and most information processing equipment is imported and is thus netted out of GDP”. Without semiconductors, the GDP boost falls to 0.2 points, a third of 0.6, by our arithmetic.
As we read the numbers, the gap between Goldman’s capex and GDP estimates is the import point the bank itself makes: most information processing equipment is imported, it said, and so is netted out of GDP, and without semiconductors the 0.6-point GDP boost falls to 0.2 points.
As we read the comparison, Goldman’s $1.4 trillion and BEA’s $1.77 trillion for equipment and software are different measures: the bank counts AI-related hardware and intellectual property, while BEA’s categories count all information processing equipment and all software.
Goldman Sachs, as reported by Reuters, 12 November 2026
“AI will remain the dominant driver of business investment growth in 2027”
Three Implications
FOR THE 2027 FORECAST Goldman forecasts US business investment growth of 8% in 2027, nearly double the 4.2% consensus, Reuters reported.
FOR THE NATIONAL ACCOUNTS Information processing equipment and software reached 35.7% of US business investment in the third quarter, the highest share since 1959, by our arithmetic on BEA data.
FOR GDP Goldman puts the direct 2027 GDP boost at 0.6 percentage points, or 0.2 points excluding semiconductors, according to Reuters.
The framework says: “This is not a continuation of the cloud capex cycle. This is a different cycle, with different physics.”
As we read it, a record share of business investment going to computers and software is that different cycle showing up in the national accounts. The framework predates Goldman’s forecast and does not discuss it.
What Is Not Established
We read Reuters’ report of the Goldman forecast in full, not Goldman’s note, so its definitions and method are not established here. Goldman’s AI-related hardware and intellectual property is the bank’s own measure; BEA’s categories count all information processing equipment and all software, whatever it is used for, and how much of their rise is AI is not established.
The third-quarter figures are BEA’s advance estimate and are subject to revision. The 2027 numbers are Goldman’s forecasts, against a 4.2% consensus. We did not contact Goldman Sachs for this piece.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Goldman forecasts annualized AI-related hardware and intellectual property spending of $1.4 trillion by the end of 2027, nearly 4% of the US economy, Reuters reported. By our arithmetic on BEA data, information processing equipment and software already supplied about 64% of the past year’s $534 billion rise in business investment, and reached 35.7% of it, the highest share since 1959.
A note on sourcing. We read Reuters’ report of 12 November on the Goldman Sachs forecast in full; we did not read Goldman’s note itself. The investment figures are the Bureau of Economic Analysis’ private fixed investment series in current dollars at seasonally adjusted annual rates, with the third quarter of 2026 from the advance estimate of 29 October, which we retrieved from FRED. The shares, changes and growth rates are our arithmetic on those figures. We did not contact Goldman Sachs.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
