This has been my fear for a long time - if most of Bitcoin's value come from "buying it to sell later", then sooner or later it will crash very hard until its value mostly will reflect its utility.
Ethereum is far worse than BTC due to its high centralization, and its utility is actually fake, because it just provides crowdfunding and marketplace for more useless tokens.
I agree about the centralization, and lots of the tokens on ETH are indeed useless, and there is also a kind of "DeFi circlejerk" where you use DEXes to trade DEX tokens Ethereum is far worse than BTC due to its high centralization, and its utility is actually fake, because it just provides crowdfunding and marketplace for more useless tokens.
However, there are some projects too which do have some utility, and businesses of any kind can crowdfund on Ethereum (if there aren't operating in some regulated market). They can of course also use Bitcoin, too (with "protocol layers" like Omni or Counterparty) but Ethereum is far more established at this time for this use case. So it's possible that in the case of a really deep crypto winter the crowdfunding scene keeps Ethereum a bit more alive. However, if Bitcoin recovers (due to finding real usage and adoption outside of speculation) my guess is that it will overtake Ethereum again.
The difference between crypto and dotcom is that the Internet technology was making constant and fast progress every year, while crypto has been stagnant for a long time. Bitcoin is not too different from Bitcoin of 10 years ago, and altcoins just invent new ways to sell premined tokens with no real use - DeFi, NFT, ICO and so on is essentially the same.
This may appear so at the first glance. But first, Bitcoin is evolving, there are bit improvements like Segwit and Taproot, and the biggest move are second layers like LN and sidechains (even if there is still no decentralized BTC sidechain, rollups on Ethereum are making progress in this field).Second, I believe the perceived progress in internet technology is also not that significant as it seems, the "progress" here has much more to do with adoption (e.g. in the realm of social media). Facebook or Instagram are technically not that different from a web forum of ca. 1996, it's basically a HTML frontend with a database containing messages and a bit of JavaScript. Of course hardware improvements have made new developments possible or better, like AI and more realistic visual stuff (videogames, ultra HD imagery). In fact, tools like ChatGPT, DALL-E etc. are the only field where I really could see "revolutionary" changes in the last 20-25 years. But in general: the web/tech business models which weren't working in the 90s mainly work now because the Internet is used by the majority of the population, not so much because of technological advancements.
If we see some more progress particularly in the layer-2 field, this could of course improve the chances for a Supercycle because it can make mass adoption easier.