Let f be a real-valued function defined on a set X. On the set F(X) of all finite subsets of X, directed with respect to the relation ⊂, the real-valued function H→supx∈Hf(x) is increasing, the real-valued function H→infx∈Hf(x) is decreasing, and we have
{supx∈Af(x)=limH∈F(X)(supx∈Hf(x)),infx∈Af(x)=limH∈F(X)(infx∈Hf(x)).
Let φ(H)=supx∈Hf(x). Clearly φ is increasing, and therefore has a limit a (no. 2, Theorem 2); and since φ(H)≤supx∈Af(x) for all H, we have a≤supx∈Af(x) (no. 2, Theorem 1). If we had a<supx∈Af(x), then there would exist x0∈X such that a<f(x0); but this is a contradiction since φ(H)≥f(x0) whenever x0∈H.
In particular, by (1), if A is any non-empty subset of R, we have
supA=H∈F(A)lim(x∈Hsupx).
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.