The failed "vortex-atoms" theory of matter by Kelvin and Tait had a profound impact on mathematics and physics. Building on the understanding of vorticity by Helmholtz, and observing stability of smoke rings, they hypothesised that elementary particles (at that time atoms) are indestructible knotted vortices in luminiferous aether: the hypothetical ideal fluid filling the universe. The vortex-atoms theory identified chemical elements as topologically different vortex knots, and matter was interpreted as bound states of these knotted vortices. This work initiated the field of knot theory in mathematics. It also influenced modern physics, where a close although incomplete analogy exists with the theory of superfluidity, which started with Onsager's and Feynman's introduction of quantum vortices. Indeed many macroscopic properties of superconductors and superfluids are indeed determined by vortex lines forming different "aggregate states", such as vortex crystals and liquids. While crucial importance of knots was understood for many physical systems in the recent years, there is no known physical realization of the central element of Kelvin theory: the stable particle-like vortex knot. Indeed, vortex loops and knots in superfluids and ordinary superconductors form as dynamical excitations and are unstable by Derrick theorem. This instability in fact dictates many of the universal macroscopic properties of superfluids. Here we show that there are superconducting states with principally different properties of the vorticity: where vortex knots are intrinsically stable. We demonstrate that such features should be realised near certain critical points, where the hydro-magneto-statics of superconducting states yields stables vortex knots which behave similar to those envisaged in Kelvin and Tait's theory of vortex-atoms in luminiferous aether.