The Economy of Bits Moved up One Layer: Capital Intensity, Value Capture, and the Compute Stack in the S&P 500, 2008–2025

Abstract

Market-capitalization concentration in the Standard & Poor’s 500 (S&P 500) rose sharply between 2016 and 2025, and the explanation usually offered is the economy of bits: negligible marginal cost, network effects, and intangible capital are held to let a few firms scale without scaling their balance sheets. This paper tests that explanation against firm-level measurement of the characteristic that it names. Year-end index membership is reconstructed from a historical constituent file, repaired for ticker renames against filing histories held by the U.S. Securities and Exchange Commission (SEC), and merged with firm fundamentals drawn from the SEC’s Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system for 2008–2025. A firm-level index of bit intensity is constructed from asset lightness, intangible share, research and development intensity, and gross margin. The concentration result survives the repair; the explanation does not. Bit intensity predicts membership of the top five percent of the capitalization distribution and of wider tails, but not of the top one percent. A Gabaix–Ibragimov rank regression finds the tail among bit-intensive firms to be thinner rather than fatter than the tail among their less bit-intensive counterparts, and the sign survives standardization within sector. Gibrat’s law of proportionate growth is not rejected, which closes the mechanical alternative. What did change over the period is the balance sheet of the firms that operate in computing: their capital intensity rose several-fold, a divergence from other large firms that is neither an artifact of the 2019 change in lease accounting nor a general size effect, and which carries them most of the way from the software sector towards the utility sector. That capital did not capture the corresponding value. Of the index share that the compute stack gained, the larger part went to the layer that designs processors and owns no fabrication plants, and the majority of that gain is attributable to a single firm.