It also helps the discussion to use reliably sourced numbers.
The cost of the Gateway tunnels themselves has been generally quoted as $11 billion, not $20+ billion. Even that number is high and as many of us have pointed out to the agencies like NJT and PANYNJ whenever they have given us a chance, the bloat comes partly from the "Christmas Tree Syndrome" wherein, just because you have a project that seems likely to get funded, everyone goes on and adds their favorite but ultimately unnecessary ornamental enhancement to it piling onto the cost.[1]
Finally they have taken that observation seriously and have started paring things back to the essentials thus reducing cost, the new estimate being $9.5 billion (See
this article).
The OP is not alone in getting the cost numbers mixed up. The entire Gateway project which indeed has been quoted to have the price tag of something between $20 and $25 billion consists of way more than just the tunnels, and includes the Portal Bridges (two of them, currently only the North one is being worked on), four tracks all the way from Bergen interlocking (i.e. west end of the tunnels) to Dock interlocking (i.e. Newark Penn Station east end, replacement of the Sawtooth Bridge where NJT's M&E Line ducks under the NEC, plus some Penn Station South enabling costs, though not the cost of it.
I have never been able to figure out whether the late add on (Schumer pet project to allegedly run fast trains to Stewart Airport) Secaucus Loop connecting the NEC to the Bergen/Main Lines cost is included in that large number or if that is another separate item. There are also some articles in Railway Age that confusingly talk about the (cancelled by Governor Christie) ARC related build out in NJ that NJT planned (e.g. the Koppers Coke Yard and track connections to it from the NEC) as if they have just been moved over to Gateway, which is not the case.
Several people who have looked at it closely seem to believe that it should cost something like $5 billion or so assuming averaged costs in the industrialized world applied to the complexity of the project. I don't know how right or wrong they are. Suffice it to say Switzerland has traditionally managed to do civil engineering projects at significantly lower cost than Germany, France or the UK. E.g. the Channel tunnel after its 80% cost overrun came in at (using present day value of the original cost of of around GBP 5 billion, and early 2019 exchange rates) $16 to $20 billion (See for example
this article) depending on which costs are counted as part of its construction (compare to $12 billion (2016) for the Gotthard Base Tunnels). Also using purchasing power parity instead of official bank exchange rates yield slightly different numbers. The Chunnel is three tubes like the proposed FinEst, of which the actual cost may or may not be what is currently estimated.
So coming back to the New Hudson Tubes or Gateway Tunnels, current estimates are running at $9.5 billion. That is still high by developed world standards, and ridiculously high by world standards. But it is being built in one of the highest cost markets in the world for which of course one gets the pleasure of paying a premium in cost. There is a very active attempt ongoing for cost containment, hopefully without undermining the quality of the product.
[1] The other reason of bloat for ARC was that NJT tried to add all sorts of additional tunnels and links to add real or imagined additional traffic to the tunnels in order to meet a bureaucratic cost per rider removed from road metric to get the project a high enough rating to make it a leading candidate for FTA new start funding. An unintended consequence I am sure of a well meaning FTA bureaucratic requirement.