I think the situation is a bit different. While some companies have used XMPP for chat, I don't recall that any had advertised prominently the notion that you could use their chat system to talk to people on a different network. This was something that geeks could do (and did), but was never a selling point for the vast majority of users.
Compare this to email, where every user expects their email client to be able to email people on different providers. If, say, GMail suddenly stopped being usable to email people without a GMail address, all users would think it was broken, not just geeks.
The trend to centralize email is a more subtle one. For instance, make email work better when emailing someone at the same domain (also because it's technically easier to handle): I have seen people trained by GMail to send huge files as attachments because it works when sending them to other GMail users. Also, spam filtering: a small email provider sending email to a large one has a higher chance of being flagged as a spammer, and has to support all requisite technologies (SPF, DKIM, SSL if you don't want the email to be marked as insecure, not being on a blacklist or residential IP, etc.).
Long sorry short, I'd have trouble imagining that a major email provider could realistically stop supporting "federation". Maybe it could happen if major providers unite to create a walled garden between them and to cut off everyone else (unrealistic), or if one provider's market share increased to near-monopoly (somewhat unlikely for now because of the long tail of institutions, companies, universities, who still want to run their own). I think the main threat to email is non-federated systems that are more convenient to use on mobile (e.g., WhatsApp and others), or social networks in general. (This may sound far-fetched, but in France I do see many companies who advertise support on Twitter and Facebook but cannot be reached by email.)
It is a sad sign of the times, you can no longer (easily) set up an email server of your own and expect emails to be delivered to everyone.
Managing O365 for a few domains I quite frequently see mail from not-so-small companies with their own servers getting stuck in MS Quarantine filtering.
My email server is running on a cheap dedicated machine, with SPF and DKIM (and SSL) correctly set up, and which is only used to send my own email (no mass mailing, etc.). In many years of using this as my email system, I recall only two cases where my email was rejected: by a university, which required manual whitelisting; and by GMail once for an incomprehensible reason. Of course greylisting (delaying for a few hours the first time) is frequent, but that's fine. So I wouldn't say the picture is as grim as you suggest.
Maybe there were more cases where my mail got lost because of excessive filtering on other providers and where I didn't notice it, but there comes a point where I don't care and I can just blame the recipient if they use a provider which silently ignores some incoming mail...
XMPP just wasn't a very good fit for the problem we needed it to solve. It is a verbose, XML-based protocol (reflecting the time in which it was conceived), and it's a mess of extensions (possibly the designers getting carried away with the possibilities of XML) instead of focusing on solving a single core problem well -- this works against the ideal of federation: what happens when you send a message taking advantage of some extension to a user that doesn't support it, or to a user that does, but through an intermediate that doesn't?
Also, it doesn't tackle spam. The concept that anyone can email anyone would probably have killed email, had it not been largely (if by no means completely) solved -- ironically, this was achieved by all but doing away with federation (email now being a oligopoly of Google, Microsoft and a few more, more or less dictating who gets to send email to whom). Finally, what I suspect was the final nail in the coffin, XMPP didn't play well with mobile (not to worry, an extension, XEP-0286, currently in state "experimental" first published in 2010 is fixing that).
EDIT: I should clarify: XMPP works quite well for private (non-federated) IM-setups. I offer this as an explanation of why XMPP isn't the interoperable email of IM across the Internet, and we instead see Facebook/WhatsApp/Slack/million other incompatible things dominating the space.
The big companies gave up on it. Google, Facebook, etc.
Everyone wants their product to be the new hotness. See the copying of stickers, voice messages, and payments from Chinese/Japanese messaging products. Or how Skype became a Snapchat clone.
Companies giving up on open standards in favor of their own walled gardens does not really provide a meaningful commentary on the open standards themselves.
XMPP itself is a bit of a mess, but it's a standard with a lot of adoption, which has its own value.
A lot of adoption? How many users does XMPP have? I'm glad that XMPP is open, but as a standard, it's useless to me if none of my friends are using services that support it.
Not sure about the actual statistics, but XMPP is not just a standard for messaging between people: I think it's also intended as a backend technology for communication behind automated services. So even though there may not be many humans using XMPP as a messaging application, there may be many people who "use it" without noticing (as in, use a website, etc., which relies on XMPP somewhere).
I don't know how one could get a good estimate of this, but it is still very widely used. Whatsapp (with custom compression and federation disabled), HipChat Server, Cisco Jabber, Android push notifications, Google Cloud Print communications, etc. are all XMPP. But yes, for personal instant messaging Google and Facebook were the big players and they are no longer on it (at least as an interface to their clients)
People mostly stayed in their walled gardens; despite the possibility to leave and explore others. And the companies running chat services decided that it wasn't in their interests to allow people out of those gardens.
I guess if xmpp had been around at the time icq kicked things off things might have been different.
I used to use third party desktop & mobile chat clients for facebook and hangouts, both via XMPP. Then I guess they realized how much user data they were missing out on by not excluding third-party apps and dropped support.
It offers a full social network solution (publication, comments, replies, contacts…) and modern IM (chatrooms, stickers, history management, synchronisation between devices).
All in real-time, multi-platforms and fully built on XMPP :)
A better question might be: what can possibly replace email as an online identity - and also - do such a better job at it that I feel inclined to switch the 100's of services over to it - a task I would not take lightly.
Compare this to email, where every user expects their email client to be able to email people on different providers. If, say, GMail suddenly stopped being usable to email people without a GMail address, all users would think it was broken, not just geeks.
The trend to centralize email is a more subtle one. For instance, make email work better when emailing someone at the same domain (also because it's technically easier to handle): I have seen people trained by GMail to send huge files as attachments because it works when sending them to other GMail users. Also, spam filtering: a small email provider sending email to a large one has a higher chance of being flagged as a spammer, and has to support all requisite technologies (SPF, DKIM, SSL if you don't want the email to be marked as insecure, not being on a blacklist or residential IP, etc.).
Long sorry short, I'd have trouble imagining that a major email provider could realistically stop supporting "federation". Maybe it could happen if major providers unite to create a walled garden between them and to cut off everyone else (unrealistic), or if one provider's market share increased to near-monopoly (somewhat unlikely for now because of the long tail of institutions, companies, universities, who still want to run their own). I think the main threat to email is non-federated systems that are more convenient to use on mobile (e.g., WhatsApp and others), or social networks in general. (This may sound far-fetched, but in France I do see many companies who advertise support on Twitter and Facebook but cannot be reached by email.)