That's why so many of these new age development tools, libraries and abstractions are such incredibly janky pieces of bloat that literally require what a few decades ago was supercomputers.
Just wondering how do you reached at the energy calculation for serving that 14k page?
For a user's access to a random web page anywhere, assuming it's not on a CDN near the user, you're looking at at ~10 routers/networks on the way involved in the connection. Did you take that into account?