31 Comments
User's avatar
Chris Wasden's avatar

Very well done. I was a very active entrepreneur during the dotcom bubble, and this is so different. I remember the crazy valuations justified by eyeballs, and the many years it took to absorb all the dark fiber from over-investment. This time, we have real companies, generating real revenues, and every chip and all power deployed as soon as it comes online, not dark fiber phenomenon this time. We may need to see ROIs that justify the level of investment, but we clearly see this as real economics and not some speculative bubble justified by crazy ideas about the future that are not based in reality. I also find the rate of change so much faster than in the dotcom era. I did some measurements of this recently, comparing adoption rates, and found that we are moving 50% faster with AI than we did with the internet. It is hard, maybe impossible, to keep up with all that is changing on a daily basis.

WJM's avatar
Feb 10Edited

“This time we have real companies generating real revenues…”. You mean like real banks involved generating real loans on REAL estate? That kind of real?

I think the AI (artificial intelligence) boom is a bubble. A significant portion of the potential market for is are not interested in it or see the pernicious possibilities from its ubiquitous use in this country. I cannot find the economic sense in these huge expenditures. As far as I am concerned, I will do anything to stay away from it. I have disabled it on my computers to the point of installing linux.

I know something about business too. The thing that surprised me the most getting all that business education and experience was the mediocrity of the thinking. Much of it is pure follow the leader in one mad rush from one bubble to another. Herd mentality, thy name is MBA.

samoan62's avatar

And all we'll have to show for it is the ability to generate meeting notes and AI p*rn.

Tom Hudak's avatar

The term "semi-reliably" in regard to AI's ability to complete tasks says it all. AI is inherently a shoddy product and should be subject to far greater regulation. I guess as usual we'll have to wait for airplanes to fall out of the sky before anything is done about it.

David Muccigrosso's avatar

RE office construction being outpaced, did your analysis account for the post-COVID hangover in vacant office buildings?

uW57&tiSUv%IUV's avatar

Yet another article gushing over the skyrocketing expenditures on AI hardware and related real estate. Obviously, if we are spending this much money on AI, it must be really great! Tech CEOs really know what they are doing!

Note the huge change in tone in this piece, at the end, where it discusses the actual results and benefits of AI. One would think that something we are spending the equivalent of the GDP of major countries on would actually show amazing results, right? But as usual, the real world results of AI are minimal or zero (or negative, as useful workflows are replaced by unproven AI applications). Meanwhile, actual human workers are being sent to unemployment offices in large numbers, along with their critical talents and experience.

Remember that LLMs work by scraping the internet, appropriating the work of paid human beings (we call this "training"), indexing the content in exotic ways, and then repeatedly figuring out the most likely next word in it's output. This is neither thinking nor inventing, but just copying and pasting prior human work in a mindless effort to produce something that looks realistic to other human beings.

This is obviously a process that is going nowhere fast, as the humans who actually discover and invent things are shut out of the creative process, and AI CPUs poison the internet with their slop (which, recall, cannot be used as training material).

CJ's avatar

A fantastic summary. First there was FRED but you're introducing a slew of other quality data sources. Thanks.

Feral Finster's avatar

TL:DR:

Heads, AI wins. Tails, taxpayers lose.

M Harley's avatar

Did you even read the piece? Why would tax payers lose? These are private companies paying their own money to build factories, data centers and chip manufacturing in the US). Each center, factory, building, increases local tax revenues. And if other tech companies get eaten, so what? Most Americans don’t work in tech!

Feral Finster's avatar

Because if the boom goes bust, the oligarchs will run screaming for a bailout.

Same as it ever was.

M Harley's avatar

lol this not true at. In most busts, the US didn’t bail people out (NFTs 2001 dot com, Enron) and the one time the US *did* do bailouts, the US taxpayer made a $15.3 billion profit while currently owning Fannie and Freddie Mae with assets approaching $500 billion

Yall just say anything without any thought. Which is a shame, because Joey puts a ton or work into creating thoughtful analysis

Feral Finster's avatar

And institutions have no oligarch owners and those owners never ever would lobby the government or Fed on behalf of the institutions they control?

Whether or why any bailout made the government a profit is irrelevant.

M Harley's avatar

I think it is relevant because that means the tax payer made money!

And which institutions? Which “oligarchs”? Tech ones? Healthcare? Climate? Just vagueness all the way down

Feral Finster's avatar

Whether the taxpayer made money is irrelevant to the folks wailing for a bailout.

samoan62's avatar

The private tech companies are not the ones paying for the AI boom. On day 2 of his administration Trump began the Stargate project to direct $500 billion to AI investment. https://en.wikipedia.org/wiki/Stargate_LLC

M Harley's avatar

Maybe you should read the sources you cite considering it says that all of the financing for star gate is coming from private companies, with SoftBank taking the lead. None of it is federal lol

Daniel's avatar

The data visualization work here is top tier. Charts that actually inform.

TROY R PETERSON's avatar

At the risk of limiting creative expression and learning, perhaps we need to start throttling or charging for prompts that are clearly a waste of resources. If they're not a waste, then the prompter will gladly pay to have it processed.

Jim's avatar

Thanks for a great piece. You pull together the data on AI spending better than anyone.

User's avatar
Comment deleted
Feb 10
Comment deleted
Ben Kerry's avatar

Problem is that the Chinese model is much more mature...

Being utterly outcompeted would certainly mean doom. (worst case scenario America will look like all the Ba'athist countries before Arab Spring, the left behind of globalization clinging to visions of past glory)