Yes, and it will be incredibly insecure unless you either convince a ton of other people to spin up nodes and act honestly in concert with you to run your network, or you build it atop Ethereum etc. Almost the only way to spin up your own is to come up with a world-changing idea, write a whitepaper and sell it to VC. So realistically people build atop Ethereum etc, and that still takes VC money because getting anything done on the famed World Computer is like booking time on a mainframe in 1976.
You don't need VC money to boot up your own HTTPS server on a traditional stack. There are lots of ways in which traditional stacks are not very censorship-resistant, and there are many ways to improve it. But at least anyone can participate on their own terms without having to either write a pitch deck and get funded, or accede to majority rule over some really important parameters of how their network operates. VC money alone is a much more pernicious influence on what will be acceptable behaviour than installing Nginx. The money required is just as heavy censorship as having to go through Visa to accept payments.
If you think "but what about email? Isn't it prohibitively expensive to run your own email server securely?", then you have fully understood the point about blockchains actually facing the same pattern of centralisation that the rest of the web does, but worse, because the alternative (your own big enough, secure enough chain) is comparatively even more difficult to adopt than the mainstream (Eth). You don't need ten million dollars and your own cryptographers and formal verification specialists to write or run an HTTP server, IMAP server, or IP network. Each of which you can do exactly as you please, essentially nobody else dictating how you do it and with costs at the bare minimum, and all of these interoperate with other people's just fine.
If you want an example of a "web3" technology that's not subject to these arguments, just look at IPFS. It's just a bunch of 20-year old ideas combined (bittorrent-esque peer discovery, mixnet routing, content-addressed blob hashes in git). None of the above problems are any worse than they are in Web 2.0, and it is leaps and bounds more censorship-resistant. You could plug in a permanent node or two today, subscribe it to a stream of authenticated tree hashes, and use it as your personal dropbox. If more people used it, you could just publish web pages there, by dragging and dropping a folder. If you have another technology that you think is a Web3 candidate, please compare it to that.