We put forward the hypothesis that the weak W boson be a compound of two 2-component Lorentz spinors. The resulting novel γWW vertex is no gauge field structure. Nevertheless, the Born amplitude of γγ→WLWL respects partial-wave unitarity. As in the Yang-Mills case, the amplitude consists of a direct term, a crossed term, and a sea-gull term, and no unobserved particles are to be involved to get the “good” high-energy behavior. This is due to an imaginary pseudoscalar γWW interaction term. Significant differences between angular distributions and total cross sections of the non-Abelian case and the case of the composite bosons are displayed. The unitarity constraint applied to the reaction γγ→WTWT leads to the prediction of the existence of a composite charged weak scalar Φ±. It constitutes the spin 0 state of the constituents forming W±. Furthermore, the existence of a second and heavy scalar-vector pair ω-X is predicted. These weak boson states are found to exclude the presence of a seagull graph. In the threshold region, the total cross section of γγ→WW in the compositeness case is smaller than in the non-Abelian case. In a broad intermediate energy region it can be larger. Upper unitarity mass-bounds are estimated. They suggest mΦ≈mw so that Φ± might be discovered by forthcoming experiments. The structure of the γΦW, γXW and γωW transition vertices can be inferred without making recourse to unitarity. However, unitarity requires that the mass relation mΦ/mW=mω/mX be valid.