Any positive integer $a$ works (e.g., $a = 3$, as suggested in Peter's comment). Consider
$$a = 2^{n}a_1, \;\; n \ge 0$$
where $a_1$ is a positive odd integer. For any odd prime $p_i$, since $2^n$ has a multiplicative inverse, we get
$$a \equiv 2^{n_i} \pmod{p_i} \, \iff \, a_1 \equiv 2^{n_i - n} \pmod{p_i}$$
Thus, we only need to only consider all odd integers $a_1$. As $a_1 = 1$ works (with $n_i = 0$ for all primes $p_i$), have $a_1 \gt 1$. Assume there are only a finite number of odd primes $p_i$, say $m$ of them, where there exists an integer $n_i$ with
$$a_1 \equiv 2^{n_i}\pmod{p_i}$$
Using the $p$-adic valuation function, among all of the odd prime factors (if any) $p_j$ of $a_1 - 1$ (since $n_i = 0$ shows these primes are a subset of the finite $p_i$ ones), let $k_1 = 1$ if there are no such primes, else have
$$k_1 = \max(\operatorname{ord}_{p_j}(a_1 - 1))$$
For an integer $k \gt k_1$ and using Euler's totient function, let
$$s = \prod_{i=1}^{m}\varphi(p_i^k)$$
Then, for each $1 \le i \le m$, we have
$$2^{s} - a_1 \equiv 1 - a_1 \not\equiv 0 \pmod{p_i^k} \;\;\to\;\; a_1 \not\equiv 2^{s} \pmod{p_i^k}$$
Thus, since $2^{s} - a_1$ can only contain at most $k_1$ of each of the prime factors among $p_i$, but $2^{s} - a_1$ is unbounded as $k$ grows, then it must have prime factors other any $p_i$ for sufficiently large $k$. However, this contradicts there are only a finite number of such primes, so there must be an infinite # of them.