Comparing AI Costs to Company Needs across S&P 500 Companies

In order to better assess the role of AI in driving market concentration, we analyze statements from 500 companies about four relevant topics: AI, Research and Development, Compute Infrastructure, and Software. We use a model to extract relevant statements between 2022 and early 2025, and outline relevant trends.

Dataset Description

• We use data from companies' earning calls and SEC 10-K filings. Quarterly earning calls are an opportunity for companies to present their strategy and competitive assets, which is a good place to find information about a company's current priorities and the investments that they consider to be the most meaningful or promising. 10-K statements are required and audited yearly filings with the SEC that follow a pre-defined format, and are a mix of tabular and narrative data. They typically become available in June or July of the following year, so the latest Fiscal Year available was 2024 at the time of writing.
• Using a Qwen3 model¹, we extract statements that relate to the topics of interest (full prompt in the appendix). We extract a total of 19,276 from earnings calls between Q1 2022 and Q1 2025, and 14,880 from 10K statements for FY 2022, 2023, and 2024 - note that the fiscal year ends in October

Trends in S&P 500 Technological Investments

Rising Importance of AI across Industries

Show data-supported trends for the sharp rise in AI interests

Reported Investments Amounts and Types

Find supporting figures for amounts invested in R&D in general and AI in particular. Compare to (comparatively low) costs of training models we know about.

Partnerships, Purchases, and Development

Compare how companies discuss investment in AI vs. generic software.