I would predict that data localization leads to data traveling shorter distances because most people access their local region rather than everything going to us-east-1.

Just use a regular data center.

The simple answer is that it's no longer a single application. It's a separate copy of the app per country/region.

should the app forward the request to the EU node (and how would it know?)

You have to add a home jurisdiction field to the account.

Everybody hypnotized themselves into believing that containers are not secure and can never be made secure so they run one container per VM. Instead of investing in making containers secure, the industry decided to invest in making VMs ligher, so VMs are now efficient enough that you can run one container per VM.

Why run Docker in VMs instead of using VM images? Because Docker's build tools are more popular than Packer-style tooling.

I think it's worth distinguishing between "they say it's not backed and it's definitely not backed" vs. "they say it's backed but it's probably not". These are different problems.

I don't think AWS has ever been low-margin and the pricing even looks deliberately complexified in some cases, like it was designed by some kind of quant.

Not really; if Vantage works as advertised it's basically free money and it should be multiplicative with any savings due to app changes.

https://varnish-cache.org/docs/trunk/phk/notes.html

Amazon Kuiper is going to compete.

I don't think US companies can legally export anything to Venezuela. Also, I'm not sure that starving people need $100/month Internet access.

I can understand why people are reluctant to cut up a $600 terminal with a Dremel.

Either SpaceX paid $250K or they didn't. The non-denials sound pretty vague.

Even if ground stations are a bottleneck now, ultimately each cell is served by one satellite and each satellite has finite capacity.

He's referring to highly speculative theories like https://discovery.ucl.ac.uk/id/eprint/10062262/

It works: https://www.youtube.com/watch?v=8PDVURcvWeg

Realistically people don't max out all their ports at once. Heck, most of the ports on X670 boards probably won't be used at all.

That's signal integrity not latency though.

Sub-ambient cooling has problems with condensation and thermal stress. With larger amounts of liquid nitrogen there are concerns about leaks causing asphyxiation. And of course there's the power needed to condense the nitrogen again.

There's some kind of market segmentation at play where nobody will make a high-quality board with a low-end "chipset" even though it would be perfect for many customers.

Won’t the timing of each chipset on each board need to be painstakingly individually configured to avoid latency from a “default” profile used in all chips (which inherently vary between every batch)?

No, because PCIe works completely differently from RAM. PCIe is a packet-based protocol that's very latency-tolerant. (Multi-lane PCIe links may require deskew but that's a standard feature that has existed all along.) I think it was Marcan who tunneled PCIe over RS-232 and the chips didn't even notice.