I don't think you quite understand the scale of the new AI data center build-outs. The new builds are requesting new total power plants/production of more than the current total installed base of power plants (in TX they have 3x the current installed base as new power) so basically they are going to double(or more) the current fuel consumption for power production of the entire US.

Because of the power consumption needs the new data centers are bringing their own baseload generators (gas turbines/diesel gens) and running them almost 100% of the time with the associated noise they produce.

The attitude is also common among the "well trained". They will say "we know this..." and often the proof is that the theory has held for some number of years so that makes it such that it must be true. And it seems that in pretty much every field greater than 50% of the experts are just "well trained" and don't really have a feel nor an actual understanding of anything. "well trained" can typically do that math but don't understand how ANY of the math relates to reality. And religion is all training...

Dark matter and expansion never made sense to me. Dark matter was invented to explain expansion, and it always seemed to me that expansion was some sort of measurement error/mis-calculation via not quite having the cosmic distance ladder quite right and because of that screwing up the distances and/or speed calculations.

People often buy by selecting for certain features, and you get stuck with some features you don't like because the perfect choice is not available.

And then clueless exec's claim that the feature that they got stuck with must be valued because they bought it (simply because they had no choice to get some other feature they though was important). This is especially true when features are bundled into large packages and/or you simply cannot pick less (or even different) electronics with a given package that has the feature that you want. And getting simple/minimal electronics then you can only get the absolute base model (if that even exists). When buying vehicles if you want one feature, that usually comes with 10+ other features/add-ons several of which you don't care about at all, and several that you may hate after a while.

Many different AI's picking many different sets of winners. One of them happening to get it right is kind of expected if there are enough AI guesses. The one that happens to get it right claims it knew what it was doing and was not simply lucky.

Compaq (HP/HPE Intel Desktop/Server predecessor) had that in the late 1990's. It was called Compaq EasyErase/EraseEaze(I think) as one side was default used as a backspace and/or delete key. It was a dumb design especially since they defaulted it to on and initially did not have a label on the back of they keyboard telling you the key sequence to switch it into its different modes. The design died sometime later. It might have been a good idea at some level, but delivering it defaulted on using either side of the spacebar is not going to make friends as touch typists tend to use one side or the other of a spacebar. I am guessing whatever ID10T default the keyboard used the other half of the spacebar (that was still a spacebar) so though it was a good idea without have the slightest clue that different people used different sides of the spacebar.

They will deliver some the data centers and hardware and then those big 3 and a few others will be stuck making payments on the bonds for hardware and data centers that they may not have any significant customers and they hopefully are profitable enough that they won't go out of business getting stuck holding the bag.

The people making the purchases seem be purchasing the capacity because they think the magical AI is going to make them lots of money, if it does not (and that seems pretty likely that it is going to fall short of what they are spending) then there are going to be a lot of bond holders holding worthless paper at multiple levels.

SpaceX is not public, it only has an estimated value and has not IPOed yet.

So while it is big deal SpaceX's value is using funny money to buy a company whose value is also funny money.

SpaceX claims it had $16b of sales and $8b of profit, so its PE ratio is 125(at $1 trillion value) and even if they were to have the entire launch market that won't increase its sales that much. It is buying a company that has an infinite PE ratio (has not made enough revenue to even pay its loan costs, and is burning cash).

Basically 2 companies that are claimed to be large on paper based on expectations that may or may not EVER actually materialize.

If you cannot live on your salary because you keep investing(ie betting) in options or gambling and losing money because these are on average no win schemes (unless you are the house) then you probably aren't smart enough to figure out stealing the phones is also a no win, and you are going to get caught.

It is not as if HP-laptops/HPE-enterprise is any more competent.

If either of them had a clue it might be trouble for the other one, but there is pretty much zero chance of that happening.

I have to think you are right. Anyone with slightly questionable credit will either have their card outright cancelled and/or simply have the limit massively reduced.

And the short term affects would be a sudden blip in consumer spending like you noted.;

Likely since he wants to do this, is that it is simply it polls good with his followers(who must on average have a lot of credit card debt so believe this will help them) that don't understand their credit card/line is going away the moment 10% is the limit.

They were only "decent" then because everyone else sucked at getting a stable db. I tested 4 db in about 1990 and the other 3 worked badly. Oracle's install script had some poor assumptions but once worked around worked as a db. One of the DB got installed and kept getting "internal errors" and support says reinstall will probably fix it, so given we were evaluating it for 50+ installs that lack of understanding was unacceptable.

And unlike banks that fail and take others money with them, Oracle failing simply gets their assets (the db, the cloud and other things) sold to someone else, assuming the other possible buyer's don't also go all in on borrowing for AI and lose their shirts also.

It seems like double/triple dipping.

Cross license with Intel/AMD and get paid per cpu sold.

Cross License with Microsoft and get paid per windows copy sold for the software decode.

And criss license with all of the OEM hardware makers and get paid yet again for each machine sold that can decode in hw.

Great for Nokia to double and triple dip and get paid 3 times for the exact same users. This really indicates we have a royally screwed up patent system.

What I don't understand about all of the half-assed patent systems, is how they are allowed to sue HP and/or Amazon when the actual violator is Intel and/or AMD including the features in the graphics chip. And I would strongly suspect that Nokia has a patent sharing agreement with both Intel/AMD and so cannot sue them, so is suing all of intel's large customers. You have to love double-dipping...

Either Autocomplete on steriods or Correlation on steriods. Every useful right answer it gives is not original and simply a mis-mash of other's work. As you say, neither is a sign of any actual intelligence. But the believers say AGI is "right" around the corner. I predict it will get here the week before the Faster-Than-Light Drive goes operational, and the week after we get commercial Fusion working...

Not a surprise they are underwater.

I have personally heard(first hand, I retired so no longer work there) multiple people in the company(including at least one of the CEOs and Larry) hype AI as being able to compress and store medical images at lower resolution because when you use AI superzoom to enlarge/up-scale the lower resolution compressed image it can recreate the "original" detail and can be used for diagnoses.

Given they think AI is magic and they seem to believe Hollywood superzoom exists they will pay anything for that unobtainium, that is not ever going to be possible to ever do (under ANY conditions) because it for any given 4 pixels you attempt to zoom into up there are mathematically a huge number of (3x3,4x4) solutions and no way to determine which solution was/is true.