A key challenge to the scalable deployment of the energy self-sustainabil-ity (ESS) Internet of Everything (IoE) for sixth-generation (6G) networks is juggling massive connectivity and high spectral effi-ciency (SE). Cell-free massive multiple-input multiple-output (CF mMIMO) is considered as a promising solution, where many wireless access points perform coherent signal process-ing to jointly serve the users. However, mas-sive connectivity and high SE are difficult to obtain at the same time because of the limited pilot resource. To solve this problem, we pro-pose a new framework for ESS IoE networks where the user activity detection (UAD) and channel estimation are decoupled. A UAD de-tector based on deep convolutional neural net-works, an initial access scheme, and a scalable power control policy are proposed to enable the practical scalable CF mMIMO implemen-tation. We derive novel and exact closed-form expressions of harvested energy and SE with maximum ratio (MR) processing. Using local partial minimum mean-square error and MR combining, simulation results prove that the proposed framework can serve more users,improve the SE performance, and achieve better user fairness for the considered ESS IoE networks.