You don't need lock-in to be anti-competitive. The requirement of extra work to implement AMP to get that higher search results page placement is the issue.
At which point pushing for new technologies as a private entity is anti-competitive vs moving technology forward?
If the criteria is just "needs extra work" then unfortunately almost nothing can change and we're all going to live with the existing technology. Change inherently has friction and requires "extra work" with the hope that's an investment which provides returns long term.
In other words, say you are a large Internet company that is trying to improve web page loading times. You profile why most web pages are slow and identify issues. You publicly report on those issues and develop guidelines and criteria. Nobody bothers because "extra work". You develop new technology that directly addresses those issues, this technology works within the existing environment but it requires both client and server support to be most effective. Do you think anyone cares? No, because of "extra work". That's why there needs to be incentives. Now you have a "penalty" for not doing that "extra work". You can file it under "it's anti-competitive" (maybe it is) but if you do the "extra work" then suddenly the anti-competitive part works for you, not against you. IMO that's why it's not anti-competitive.
Other examples: why do you think there are so many people that complained when iPhone released with Flash reader? "extra work". Similarly when it removed the audio jack. Change is friction and friction is extra work. But most of the time that's not anti-competitive...
Let me explain based on my 15 years of adtech experience:
HTML is already fast (see HN for an example). HTML is already universal across devices and browsers. HTML is already published and easily cached for billions of pieces of content.
AMP is a fork of HTML that only targets mobile browsers specifically linking from Google search results. It's useless on its own, but AMP is required to get higher placement on search results pages, so publishers are effectively forced to spend technical resources to output an entirely new format just to maintain that ranking.
If Google wanted faster pages then it can do what it always does and incentivize that behavior by ranking results based on loading speed. These signals are already collected and available in your Google webmaster console. There's nothing new to build, just tweak ranking calculation based on existing data. Sites would get faster overnight, and they would be faster for every single user because HTML is universal.
Do you know why they didn't do that? Because it's the ads and tracking that's slow, not the HTML. Google's Doubleclick and Google Analytics are the biggest adserver and tracking systems used on the web. This entire AMP project is created to circumvent their own slow systems. It creates more work for publishers while increasing data collection by running a "free" CDN that never leaves a Google-owned domain and thereby always supports first-party cookies. It's a perfect solution to protect against anti-tracking browsers and why Chrome now will also block 3rd-party cookies, because it won't affect AMP hosted on Google domains.
First party storage won't be affected without some major AI tech in browsers so cookies are still the best deterministic connection, especially since most people are logged into a Google service already (gmail, chrome, android, youtube, etc).
Probabilistic techniques are used for anonymous users or environments like iOS Safari that are very strict.