\contentsline {section}{\numberline {1}Introduction}{4}{section.1}%
\contentsline {paragraph}{Contributions.}{5}{paragraph*.1}%
\contentsline {paragraph}{Caveats up front.}{5}{paragraph*.2}%
\contentsline {section}{\numberline {2}Related Literature}{6}{section.2}%
\contentsline {paragraph}{Text as data.}{6}{paragraph*.3}%
\contentsline {paragraph}{LLMs for extraction and measurement.}{6}{paragraph*.4}%
\contentsline {paragraph}{Corporate investment.}{6}{paragraph*.5}%
\contentsline {paragraph}{Innovation and diffusion.}{6}{paragraph*.6}%
\contentsline {paragraph}{Networks and representations.}{7}{paragraph*.7}%
\contentsline {section}{\numberline {3}Data and Extraction Pipeline}{7}{section.3}%
\contentsline {subsection}{\numberline {3.1}Source corpus}{7}{subsection.3.1}%
\contentsline {subsection}{\numberline {3.2}LLM extraction}{7}{subsection.3.2}%
\contentsline {subsection}{\numberline {3.3}Resolution, graph, and embeddings}{7}{subsection.3.3}%
\contentsline {subsection}{\numberline {3.4}The economics of extraction}{8}{subsection.3.4}%
\contentsline {subsection}{\numberline {3.5}Analysis datasets}{8}{subsection.3.5}%
\contentsline {section}{\numberline {4}Descriptive Evidence and Stylized Facts}{9}{section.4}%
\contentsline {paragraph}{Volume and trend.}{9}{paragraph*.9}%
\contentsline {paragraph}{Composition.}{9}{paragraph*.11}%
\contentsline {paragraph}{Sectoral structure.}{11}{paragraph*.14}%
\contentsline {paragraph}{Capital and geography.}{12}{paragraph*.17}%
\contentsline {subsection}{\numberline {4.1}Six stylized facts}{13}{subsection.4.1}%
\contentsline {section}{\numberline {5}Econometric Framework}{13}{section.5}%
\contentsline {paragraph}{(i) Count models.}{14}{paragraph*.19}%
\contentsline {paragraph}{(ii) Hedonic capital regressions.}{14}{paragraph*.20}%
\contentsline {paragraph}{(iii) Diffusion.}{14}{paragraph*.21}%
\contentsline {paragraph}{(iv) Difference-in-differences.}{14}{paragraph*.22}%
\contentsline {paragraph}{(v) Survival.}{14}{paragraph*.23}%
\contentsline {paragraph}{(vi) Networks and representations.}{14}{paragraph*.24}%
\contentsline {paragraph}{(vii) Prediction.}{15}{paragraph*.25}%
\contentsline {section}{\numberline {6}The Extensive Margin: Count Models}{15}{section.6}%
\contentsline {section}{\numberline {7}The Intensive Margin: What Drives Announced Capital?}{16}{section.7}%
\contentsline {section}{\numberline {8}Technology Diffusion: The Rise of Corporate AI}{18}{section.8}%
\contentsline {section}{\numberline {9}COVID-19 and Digitalization: A Difference-in-Differences Test}{19}{section.9}%
\contentsline {section}{\numberline {10}Project Lifecycles: Survival Analysis}{20}{section.10}%
\contentsline {section}{\numberline {11}The Innovation Network}{22}{section.11}%
\contentsline {section}{\numberline {12}Latent Themes: Embedding the Project Corpus}{24}{section.12}%
\contentsline {section}{\numberline {13}Predicting Mega-Projects}{25}{section.13}%
\contentsline {section}{\numberline {14}Robustness and Validation}{26}{section.14}%
\contentsline {paragraph}{Extraction quality.}{26}{paragraph*.44}%
\contentsline {paragraph}{Estimator robustness.}{27}{paragraph*.45}%
\contentsline {paragraph}{Parallel trends.}{27}{paragraph*.46}%
\contentsline {paragraph}{Level vs.\ composition.}{27}{paragraph*.47}%
\contentsline {paragraph}{Sample.}{27}{paragraph*.48}%
\contentsline {section}{\numberline {15}Limitations}{27}{section.15}%
\contentsline {section}{\numberline {16}Discussion and Implications}{28}{section.16}%
\contentsline {paragraph}{For methodology.}{28}{paragraph*.49}%
\contentsline {paragraph}{For the study of investment.}{28}{paragraph*.50}%
\contentsline {paragraph}{For the AI-diffusion debate.}{28}{paragraph*.51}%
\contentsline {paragraph}{For practice.}{28}{paragraph*.52}%
\contentsline {section}{\numberline {17}Conclusion}{29}{section.17}%
\contentsline {section}{\numberline {A}Project Taxonomy and Variable Definitions}{30}{appendix.A}%
\contentsline {section}{\numberline {B}Pipeline and Reproducibility}{30}{appendix.B}%