Does the concept of General Purpose Technologies help explain periods of faster and slower productivity advance in economies? The paper develops a new comparative data set on the usage of electricity in the manufacturing sectors of the USA, Britain, France, Germany and Japan and proceeds to evaluate the hypothesis of a productivity bonus as postulated by many existing GPT models. Using the case of the diffusion of electrical power in the early twentieth century this paper shows that there was no generalized productivity boost from electrical power diffusion as postulated by many existing GPT models. The productivity gains from this GPT varied widely across economies and industries, suggesting that the power of GPTs to predict aggregate or sectoral growth is limited.