Let me answer the basic question first and then I'll push back on few points.

In short, there is no additional benefit as you describe it.

if I go through all that effort I may as well make my site full PWA with a mobile experience.

The added benefit is when you combined AMP with the PWA thus creating a fast and seamless interface between the user and the app.

Now the push back...

AdSense revenue on amp mobile is terrible

Yes, simply looking at your stats may suggest that this statement is true, but you need to compare apples with apples. AMP pages are only ever accessed on the first page load, that is on the landing page from search. Once on an AMP page any further page views take the user to a non-AMP page. For AMP ads to get clicks requires that the user lands on a page and then clicks an ad without any further interaction. I would guess that the CTR of AMP pages is thus much lower than non-AMP pages, but if you compare the AMP-ad CTR specifically to the CTR of non-AMP ads on landing pages only that the CTR and the resulting RPM would similar.

AMP is supposed to be a gateway page. AMP pages may be providing added benefit that you do not realize. In a worst case eliminating the AMP pages may reduce traffic to your site, thus reducing the total number of non-amp ad clicks. Replacing the AMP ads with non AMP will not likely provide much of a benefit as the poor performance has likely more to do with the fact that the ads only appear on landing pages.

I have a mobile experience outside of AMP, so I would still serve correct & fast mobile experience, just not constrained by AMP... this is a WordPress site where AMP operates as its own theme and with very tight constraints. They're not landing pages and if they were, the experience of linking to the non-amp page for everything of value seems incohesive and sort of makes me want to stick to the mobile experience I have already.

I'm just at the point not wanting to bother with changing amp rules, another theme/template to manage or custom content for user experience that shouldn't be different but is different because of the limitations of AMP with regards to dynamic content. (forums/community/interactive sites). My experiment is done i guess i could say :)

So how would removing AMP pages reduce traffic? is AMP pages indexed in a separate AMP only Mobile index? By purging the AMP cache and removing the amphtml element, it should just default to my awesome amazing mobile experience anyway should it not?

BTW, one of my advertising partners has said that their RPM is about 80% below market value on AMP because advertisers aren't paying near full rate for AMP clicks.

In many ways, I can't wait to run my standard google tag manager and get back to tracking what I need to track to be successful that becomes impossibly difficult or constrained while running amp.

Also, why is bounce rate for AMP so high? I've asked other publishers that ran amp their bounce rate was north of 90-95% but when removing amp it went to their normal rate. Seems AMP is used in framed browsers/apps and the shole purpose of AMP is to make that easier so people "buy an app" or "use an app" for that - which sucks.

news sites who want in the carousel on google search probably do well, but i'm not featured there...

They're not landing pages

Then having an AMP version of such a page is completely and utterly pointless.

AMP and all it constraints are there to ensure that the pages can be cached safely on a (or many) server(s) that is(are) not related to the domain and that those cached pages can be served with out errors to the user regardless of their browser. If your are not taking advantage of that caching, then there is no point. Note that this isn't the same as caching assets on a CDN. It is caching the page in such a state that there is nothing to block its rendering.

If all you are doing is serving AMP pages to mobile users that are already on your site then you have been gravely misled as to the purpose of AMP. Get rid of it.

BTW, one of my advertising partners has said that their RPM is about 80% below market value on AMP because advertisers aren't paying near full rate for AMP clicks.

That makes perfect sense to me. Why would an advertiser want to pay full price for traffic in which there is a much higher proportion of users for which they are unsure of the degree of engagement. Deep content that garners engagement will always earn more. The issue is not the framework, it is the type of user.

Typically there is very little tracking by publisher of engagement metrics at the level of the specific page for which their ads appear. Typically, they place some ad units across many pages and then watch the main dashboard for the aggregate metrics. Given that AMP forces publishers to use specific ad units it cause the "bias" described above to be revealed. The publishers sees the biased metric and compares them to all the ad-impressions that are now equally but inversely biased. If a website has a big percentage of AMP traffic, most of the users that bounce will bounce on an AMP. Thus, pages that are not AMP will have fewer user that will bounce and thus users are guaranteed to be more engaged. You are comparing apples to oranges without realizing it.

So how would removing AMP pages reduce traffic? is AMP pages indexed in a separate AMP only Mobile index?

This should be of no importance to you since the pages in question are not landing pages.

Also, why is bounce rate for AMP so high?

I think it has to do with how AMP is treated in GA. AMP pages are not served from your server, they are served from the CDN, I assume that it depends on the user's browser but I believe that when the user navigates to a link on your site that this action is sometimes counted as a bounce because the user is switching domains. I know that in real-time I often see users as appearing twice in the same location once for an AMP page and the other for a non-amp page that is directly and only linked to that AMP page. This inflates the bounce-rate of AMP pages and the number of users in general (as the users is counted twice).

I enabled AMP as soon as that "opportunity" came out, for more than 3,000 pages on my main website.
Now, I am seriously considering disabling AMP completely, and forever. Why?

Not because of the Adsense revenues AMP pages are bringing to me, CPC is abysmal, but CTR is very good, so advertising revenues are not really a problem.

The REAL problem is that AMP is redundant. It costs us a lot of time and money to comply with all those ever-changing validation rules. I've enabled a well-reputed GDPR plugin for WP and...boom. 3,000 page suddenly did not validate in GSC (and that's obvious, no way an AMP page including Adsense can really comply with GDPR, unless you completely tweak the AMP and Adsense code which, AFAIK, we are not allowed to do).
I tried to update the "official" WP plugin for AMP and...boom, I found out WP 4.8 was not supported, and I was required to upgrade to WP 5.1, something I don't need and don't want either.
I've also found that, in no way, AMP pages produce a better user experience when compared to reasonably-optimized standard mobile pages, period. Many times, it's just the opposite, with AMP pages lacking some pictures or loading much slower than the original, non-AMP, ones.

Furthermore. Traffic from AMP? Marginal, though I am into that infamous "Top Stories" carousel. Quality of traffic from Google via AMP? Horrible. The worst time on page, bounce rate, and conversion rate I've ever seen for 20 years.

So, what's the point of enabling AMP in your website? Ohhh, I know the answer... Because Google wants you to do so, and we are all frightened as hell to ruin our ranking by removing those AMP pages which cost us blood, sweat, and tears and pays us just peanuts....

And, maybe, that's just another "SEO myth" and Google simply doesn't care whether your pages are AMP or not, as long as they are reasonably fast-loading and clean.

OK, I see a lot of hate for AMP in this thread and I realize some of it is justified.

For those of you who ran tests, well done. Never take anything at face value, always test (if possible) just because something is new and "proven" (proven by who?) doesn't mean it is going to work for you or that it is even good .

For people who claim their websites are already fast, I see some mentioned 1.5 seconds - is that in a 3G network or your home PC super fast internet? AMP was made to be specifically fast for 3G in which site speed should be tested. Also, the average site load of AMP sites is 0.5 seconds - [ampproject.org...] 3 times faster than 1.5 seconds, that's for sure.

ByronM, you said it will take a long time to redesign your site in AMP due to the client site libraries, etc - do you REALLY need them? The reason AMP is so fast is that it makes you second guess your choices - just cause you have that super fancy pop up, doesn't mean you need it or it is useful, but it is a drain on the user's PC/phone.

riccarbi, as for AMP being a "google myth" that's what I initially thought, but Google something and see most of the 1st page results are AMP pages.

I was very skeptical of AMP too, but the truth is most of the websites are developed horribly. However, you can have an AMP site and it will still be horrible :-)

I never went live with any AMP page either. I serve static pages and implement any interactive stuff from 3rd party platforms like Disqus. I don't use wordpress, I've sort of built my own custom publishing back end over the years that requires no database, pages are in the 99th percentile fast.

For my site adding AMP slows down that first page load so it's a whole lot of added headache for a net loss in speed.

The recent core algo seems to have punished me a bit for not having AMP.

It's a lose-lose situation for me, and for many, right now. My advice is that if you're starting out a new site add AMP from the start but if you have an old site with solid history and good backlinks then the pages can take a bit more of a punch before falling off, AMP is optional if your pages are fast and aged, IMO.

I won't be adding it unless I can't find a way to rank in the 99th percentile for speed without it. Speed is important but AMP isn't the only way to get fast, even if Google suggests it is.