FRED: Business R&D Expenditure – Information Industry
- Rows
- 318
- Columns
- 2
- Data range
- 1947-01-01 to 2026-04-01
- Auto-refresh
- Re-checks its source every 14 days
- Last refreshed with new data
- 2026-10-11
Quarterly R&D expenditure in the information sector, a leading indicator of AI investment cycles including compute and model development spending.
AI analysis
Dataset Analysis: FRED Business R&D Expenditure – Information Industry
What This Dataset Represents
This dataset captures quarterly R&D expenditure in the U.S. information sector, sourced directly from the Federal Reserve Bank of St. Louis (FRED) via their public API at fred.stlouisfed.org. FRED is a highly authoritative, regularly updated macroeconomic data repository maintained by the St. Louis Fed, lending strong credibility and freshness to this series. The variable Y694RC1Q027SBEA tracks spending in billions of dollars and spans an exceptionally long historical window — from Q1 1947 through Q1 2026 — making it one of the more longitudinally rich indicators available for studying the evolution of technology investment. Its designation as an AI indicator is well-reasoned: information-sector R&D expenditure acts as a leading proxy for AI investment cycles, capturing upstream spending on compute infrastructure, model development, and software research before that spending manifests in commercial AI products or labor market shifts.
---
Data Quality Observations
Data quality here is notably strong. Across 317 rows and 2 columns, there are zero null cells, meaning both the date index and the expenditure values are fully populated — a relatively rare quality for a long historical series that spans nearly 80 years. There are no reported type mismatches, and the date column carries 317 distinct values, confirming a perfectly clean quarterly cadence with no duplicated periods. The expenditure column shows 316 distinct values (out of 317 rows), indicating at most one repeated value — effectively negligible and likely coincidental rather than a data integrity issue. The absence of duplicate rows and nulls means this dataset can be used in joins and time-series models without preprocessing for missingness, which is a meaningful advantage.
---
Key Column Distributions
The expenditure column (Y694RC1Q027SBEA) tells a compelling growth story through its statistics. The mean of \$234.5B substantially exceeds the median of \$132.5B, a divergence driven by the heavy right-skew (skewness = 1.47) characteristic of exponential growth series. The interquartile range spans \$24.0B (Q1) to \$365.7B (Q3), reflecting a roughly 15x spread between the lower and upper halves of the distribution — consistent with a dataset that starts at near-zero spending in the postwar era and accelerates sharply in recent decades. The minimum of \$2.4B and maximum of \$1,143.2B bookend that arc. The 17 flagged outliers likely cluster in the most recent quarters (post-2020), where AI-driven investment surged dramatically, and these outliers are analytically meaningful rather than erroneous — they represent the signal, not noise.
---
Recommended Join Key
The observationdate column is the clear and only join key, and it is well-suited for that role: zero nulls, 317 distinct quarterly timestamps, and a consistent ISO date format spanning 1947–2026. For cross-dataset joins, quarterly granularity should be the alignment target. Datasets with monthly or annual frequency can be aggregated or interpolated to quarterly, but this series should not be further disaggregated without auxiliary data. Date range alignment will be the primary practical constraint — many AI-era datasets will only overlap with the post-2010 or post-2015 portion of this series.
---
Suggested Pairing Datasets for Correlation Analysis
Several dataset categories would pair powerfully with this series:
- AI Patent Filings & Publications (USPTO, Google Patents, Semantic Scholar): correlating R&D spend leads with IP output would reveal investment-to-innovation lag structures. - FRED: Private Nonresidential Fixed Investment in Software or IT Equipment — to decompose whether R&D spending co-moves with or leads capital deployment. - NVIDIA / Semiconductor Revenue or GPU Shipment Data: compute procurement is a near-real-time proxy for model training spend, and a lag analysis against this R&D series could be highly revealing. - BLS Occupational Employment in Computer & Mathematical Roles: testing whether R&D expenditure predicts AI-adjacent hiring with a 2–4 quarter lag. - S&P 500 Information Technology Sector Index or VC AI Investment Volumes: to explore whether public/private market valuations lead or lag fundamental R&D commitments.
The long historical window also makes this dataset uniquely valuable for regime-change analysis — distinguishing the dot-com era, mobile/cloud era, and generative AI era as structurally distinct R&D investment regimes.
Columns
- observation_date (date)
- Y694RC1Q027SBEA (decimal)