Here are some not obvious problems with big blocks:

• they take longer to propagate. This means that miners that are closer together or have better networking have a big advantage. You mentioned DC networking. If miners have to run in datacenters, you end up with s smaller number of very centralized miners. Easy to capture. Might as well just use a bank.
• the only thing that keeps miners from colluding to change consensus rules (spending coins they don’t own, changing the supply schedule, mining more than 21M coins) is users running their own nodes. What does the initial block download (IBD) look like if you have 8GB blocks? It becomes impossible for anyone to validate the chain for themselves, so everyone just uses SPV clients which check for tx inclusion but doesn’t validate the whole block.

These ideas that your friend has are not unique. They’re basically the position of BSV. You can take a lot at the number of full node on bsv (and count them on two hands ) and you can look at what the market thinks of bsv by looking at its marketcap. Giant blocks don’t work. Layered scaling is the only realistic approach.