You have this wrong. If you have 51% of the mining power, you do indeed have a negligible chance of winning seven consecutive blocks in a row if you mine the normal way. But, if you just mine seven sequential blocks without publishing the first six until you've mined all seven, you have a rather good chance of getting all seven blocks before the rest of the network mines seven blocks.

The idea that it would cost billions of dollars worth of ASICs to get this chance relies on the fact that Bitcoin is in an absurd boom right now, such that only new ASICs are competitive and there isn't a pile of old used ASICs available. If Bitcoin ever stagnates, this will rapidly stop being true.

You're also assuming that the ability to rewrite an hour of Bitcoin history isn't worth a couple billion dollars to anyone. That's an interesting assumption. (If I were, say, Israel, one of my enemies started using Bitcoin to buy weapons, and I thought that the ability to rewrite Bitcoin history an hour at a time could bankrupt them, I would be thrilled at the opportunity to spend a billion dollars to do so.)

I'm frankly rather surprised that Bitcoin, and proof-of-work blockchains in general, are considered secure enough for serious high-value work.