Abstract We forecast the number of galaxy clusters that can be detected via the thermal Sunyaev–Zel’dovich (tSZ) signals by future cosmic microwave background (CMB) experiments, primarily the wide area survey of the CMB-S4 experiment but also CMB-S4's smaller de-lensing survey and the proposed CMB-HD experiment. We predict that CMB-S4 will detect 75,000 clusters with its wide survey of f _sky = 50% and 14,000 clusters with its deep survey of f _sky = 3%. Of these, approximately 1350 clusters will be at z ≥ 2, a regime that is difficult to probe by optical or X-ray surveys. We assume CMB-HD will survey the same sky as the S4-Wide, and find that CMB-HD will detect three times more overall and an order of magnitude more z ≥ 2 clusters than CMB-S4. These results include galactic and extragalactic foregrounds along with atmospheric and instrumental noise. Using CMB-cluster lensing to calibrate the cluster tSZ–mass scaling relation, we combine cluster counts with primary CMB to obtain cosmological constraints for a two-parameter extension of the standard model (ΛCDM + ∑m _ν + w ₀). In addition to constraining σ(w ₀) to ≲1%, we find that both surveys can enable a ∼2.5–4.5σ detection of ∑m _ν , substantially strengthening CMB-only constraints. We also study the evolution of the intracluster medium by modeling the cluster virialization v(z) and find tight constraints from CMB-S4, with further factors of three to four improvement for CMB-HD.