But, since you asked...
It will depend largely on what's going to happen to the Bitcoin network Difficulty and how you would be mining; alone or in a pool.
If you're doing it alone, then there's only 2 variables to worry about:
The Difficulty and your Hashrate.
With those two numbers, you can calculate how long, on average, it would take your rig to find a block. A block is 25BTC, but since you're doing this alone, that's your only option (you can throw the extra 24 BTC my way, if you'd like).
That calculation goes something like this: time(s) = 2^256 / (((2^224 - 2^208) / Difficulty) * Hashrate)
It's effectively the odds of a random guess being below a certain threshold, and taking into account just how often the hardware can make those random guesses.
But the approximation will do: time(s) = Difficulty * 2^32 / Hashrate.
The current difficulty is at 29,829,733,124, so plug it in for the two setups...
2xBlock Erupter USB @ 672Mhash/s: time = 29,829,733,124 * 2^32 / 672,000,000 = 190,651,381,273.792 seconds or ~6,129 years.
1xAntMiner U2 @ 2Ghash/s: time = 29,829,733,124 * 2^32 / 2,000,000,000 = 64,058,864,107.994 seconds or ~2,059 years.
Now, do keep in mind that this is the average time. It's entirely possible that your great^8 grandchildren still wouldn't see those rigs get a block, but it's also possible that a split second after you starting them up, they happen to find a block. It's just extremely unlikely.
It also tends to get worse with time. The next Difficulty (due in about a week) seems to be hovering around 35,000,000,000, and those average intervals jump to 7,191 and 2,416 years respectively.
For this very reason, people started to use pools. Sure, they'll have to share their block reward with others, but at least with all those others pooling in, the odds that at least one of them finds a block increases rapidly.
Now, pool payout systems are well beyond the scope of this post. Let's say they're still using the naive PROPortional system, which means that your payout will be equal to your hash rate share in the group. In that case, you could do some full calculations (you would basically add a single new variable: the Pool's total hash rate - let's call that Poolrate, and then some optional variables regarding fees and such).
The calculations then become:
blocktime(s) = Difficulty * 2^32 / Poolrate.
Payout = Hashrate / Poolrate * 25.
Yourtimeuntil1BTC = 1 / Payout * blocktime(s).
So let's say you join, oh I dunno, P2Pool. Their Pool hash rate is currently about 3,500 Thash/s.
Their blocktime(s) at current Difficulty is thus about 10 hours and 10 minutes (their own site states "10h16m", close enough).
Your Hashrate / Poolrate ratio for your two setups is:
2xBlock Erupter USB @ 672Mhash/s: 672,000,000 / 3,500,000,000,000,000 = 0.000000192
1xAntMiner U2 @ 2Ghash/s: 0.000000571
So if the block reward is 25BTC, you would get:
2xBlock Erupter USB @ 672Mhash/s: 0.000000192 * 25BTC = 0.000014286BTC
1xAntMiner U2 @ 2Ghash/s: 0.000000192 * 25BTC = 0.0000048BTC
For you to get 1BTC, you'd have to mine at the pool for:
2xBlock Erupter USB @ 672Mhash/s: 1BTC / 0.000014286BTC * 10 hours = 245 years.
1xAntMiner U2 @ 2Ghash/s: 1BTC / 0.0000048BTC * 10 hours = 82 years.
( Again, subject to Difficulty changes and based on averages. The pool could find a whole bunch of blocks in a row, too.. it's just extremely unlikely. Oh, and most pools charge a percentage... and don't use PROP. And, and, and... it just gets worse, really. )
Or you could skip all that and simply divide the times by the block reward, currently 25BTC:
2xBlock Erupter USB @ 672Mhash/s: ~6,129 years / 25 = ~245 years.
1xAntMiner U2 @ 2Ghash/s: ~2,059 years / 25 = ~82 years.
So, you were already told that this is a bad (monetary) investment, based on how long it would take and how much you would expend in energy costs, and hopefully now you also see the cold hard facts of why that is the case.
( Perhaps somebody else will explain to you how to write more descriptive thread titles )
But, if you're still enamored by USB stick miners anyway, head on over to that StickMiners thread.