Historically speaking, your statement is not well supported. Essentially all cryptosystems proposed before 1970 have been cracked. Most cryptosystems proposed since then have been, too. Any particular cryptosystem might turn out to have a hard and usefully high lower bound on the difficulty of cracking it, but we can't even prove P ≠ NP yet, so we aren't even close to proving such lower bounds on cryptosystems.
Large amounts of money can and do buy the time of legions of sharp mathematicians who devote their lives to devising and cracking new cryptosystems. There are good reasons to suspect that this no longer gives the US government the kind of advantage that it used to enjoy, but hard evidence of that is also hard to come by.
There's also the possibility of using large amounts of money to trick someone into using weak encryption, a policy which was very successful with Crypto AG.
Large amounts of money can and do buy the time of legions of sharp mathematicians who devote their lives to devising and cracking new cryptosystems. There are good reasons to suspect that this no longer gives the US government the kind of advantage that it used to enjoy, but hard evidence of that is also hard to come by.
There's also the possibility of using large amounts of money to trick someone into using weak encryption, a policy which was very successful with Crypto AG.