what does this have to do with the topic? These systems run on a modified FreeBSD kernel...

You are correct on this, it isn't your English skills in question, but the writer of this content.

If only for Fancy Zones on an ultra-wide monitor, it is useful.

They were effectively held captive with no money in a city they had no contacts in, and no way home. They were threatened to basically be dropped off on the street with nowhere to go, WHILE being promised terms that effectively said "nobody will ever know you did this". Don't victim shame here, this has gone to court, and it was upheld that they were doing something illegal.

One issue with testing failures in a cloud environment is that you need to be able to simulate failures of the cloud provider itself. It may seem odd, but each service needs an option to literally "break" it, i.e. say to trigger a hang on the database, or similar. Without this, it won't be possible to fully simulate all the various failure modes that could happen, and realistically, it means that you can never be sure your DR plans will work.

So, if you are a software vendor, and want to publish on GCP's marketplace, you have to jump through all kinds of hoops to document any and all open source code you use, including ensuring that code under many license types are included in the OS image you are deploying, even when not required by the license. Next, you have to take SCREENSHOTS of where all the license information is on our image, and paste it into a spreadsheet that you submit for open source review. It is the most antiquated process I have ever seen a tech company use.

I suspect that one of the reasons why new x64 chip designs are not faster than they are is due to the desire to use hyperthreading, which complicates the designs. Yes, it may allow you to leverage more of the circuits more of the time on a given core, but at the sacrifice that the fastest single thread performance is impacted. Might it be time to remove hyperthreading from cores, or possibly, craft a chiplet design with four cores without and four cores with. High priority threads get put on the dedicated cores, and lower can run on the shared cores. Win Win

Who needs to load a plane up with gold like a fleeing dictator when you can just arrange the FBI to transfer bitcoins somewhere for you?

Scheduling conference calls is bad enough as it is. Now we have to adjust times every month? Screw that!

If this is true, I suspect that the reality is that what will happen is that graphics will have to adjust to rendering the vector graphics directly, i.e. instructions will be made to draw a triangle between three points, and fill it with a given color. How this will be accomplished at the rendering layer, I don't know, but it won't involve translating to pixels anymore.

comparing a 10 core chip with one with 24-64 cores? Really?

Before calling out "government stupidity" which is generally the mantra of those promoting capitalistic solutions, think of the 737 Max. Any large organization can make mistakes.

Really, this is business as usual for all cloud providers.

Paying ARM to do the development is cheaper than developing your own design. People pay the license costs because it is SAVING money, not costing them more. And not having to deal with the infighting as one company wants something added to RISC-V while another wants to do something else was probably a blessing.

Honestly, I could easily have seen the license holders all kicking in a few bucks each, and done a counter offer against Nvidia, with a board of directors voted in by the license holders at large. They could have structured it so they have majority control, IPO'ed it, and possibly make some money in the long run.