> But they seemingly have no choice because everyone else is doing it.

doesn't sound like valid reason

That's how arms races work though. Everyone spends a ton of money to end up no further ahead compared to their rivals and peers.

"The share price will tank if we're not seen as AI-forward" is a legitimate fear for many CEOs. It's the only logical explanation for tokenmaxxing and other behavior that only makes sense from a game theoretical perspective.

Tokenmaxxing was the management buying the marketing that AGI was here and if employees would only use AI they would herald their own replacement?

If that were true then management could've spent some time coming up with better metrics than "tokens burned". They could just halve deadlines for every project, as an extreme example.

I thought tokenmaxxing was their proxy for this: more tokens burnt = more productivity = more projects completed.

You seem to assume competence here?

big if true

Clearly true though. “Not reveryone” is doing it. Most of the bubbly(?) trillions are.

In other words, those who are doing it are winning and those who are not doing it are losing.

"A strange game. The only winning move is not to play." -WOPR

I think the more spelled out version of this is "our existing debt is tied to our equity in some way so we have to maintain our unicorn valuation, and the unicorns are building AI infrastructure".

If they admit the current pace is unsustainable, they die anyway. Might as well keep building and praying for a hail Mary.

I think you're taking off in the wrong direction.

Retirement savings are 401ks in the US and they're managed by _NOT_ the retirees, but random middle managers. Those are the people who are chasing whatever the bubble is producing because their "due diligence" requires them to be bootlickers.

The same can be said of many of the tech moguls who've sworn fealty to trump; if they didn't lick the boot, they could be considered acting against shareholders value.

So, like most things in America: the people are not in control of their actual power. Not in retirement funds, not in healthcare, and not in education. They've all been turned into bizare economic dependent indirect funding programs.

I despise target retirement age funds for exactly this reason and I have all my 401k funding in major index funds that I do not care about beyond that on a day to day basis.

The major target retirement date funds are just "funds of funds" that are distributed across stock and bond index funds in an age appropriate allocation. e.g. https://investor.vanguard.com/investment-products/mutual-fun... They're generally not actively managed funds chasing bubbles.

Article is about pension funds.

401ks are self-managed, though many stick with the default investments.

The article is not strictly about pension funds. There are several mentions of individual retirement accounts.

41K’s can be self-managed. But you have to take it upon yourself to do so if you don’t, you will get dumped into a hodgepodge fund where you have the high performing and the low performing thrown together, whose purpose is to average everything out.

I didn't say anything about the management of retirement savings.

The author claims two things:

1. AI company valuations dropped because of fears that Democrats would restrict data center construction.

2. The AI industry is doomed because of unsustainable spending on data centers.

If 1 is true then investors clearly believe that the current level of spending is appropriate and any restrictions on it are harmful to long term returns.

If 2 is true and unsustainable spending continues despite a lack of RoI, then "arms race" is the only explanation for the irrationality. It is driven by the fear that if you don't spend big, you'll be part of the permanent underclass of companies.

This in turn implies a data center construction moratorium is a blessing in disguise. Everyone has an ironclad excuse to reduce financially-irresponsible but mania-driven data center investments.

Bottomline:

"Data center restrictions are bad for the industry"

and

"Data center spending is unsustainable"

are polar opposite viewpoints and imply vastly different outcomes of the current buildout.

I don't know which one is true. And like most things, I suspect the truth is in the middle. But they can't be simultaneously true.

I think 1. and 2. could both be true at the same time.

1. is claimed to have happened because INVESTORS in AI companies rerated their investment due to a threat of construction restriction 2. is a prediction about the future of how the industry will turn out

Someone who believes (2) is less likely to have been an investor to begin with and therefore didn't participate in the price rerating for (1)

I do agree with your point that a reduction in spending would have a positive outcome overall if we turn out to be in an "arms race" with a lack of ROI.

You're right 1 and 2 are different groups of people.

Data centers are profitable. Even before ChatGPT came out the hyperscalers were having capacity issues. Non-AI growth is explosive since 2020. Cloud customers complain they can't spin up the resources they want to spin up in the cloud, and again, not just GPU SKU's.

Chinese labs are not going to care about whatever anti-datacenter measures you propose. It'd take an unprecedented effort of international collaboration to slow down AI training.

(Before anyone mentions nuclear, the amount of nuclear warheads grew a lot after the non-proliferation treaty. Historically that was how collaborative big countries were.)

Wouldn't slowing down AI training in the United States insulate the retirement investments of U.S. citizens from the LLM market, regardless of what happens abroad?

I'm not proposing anything. The article is.

The non-proliferation treaty was mostly to prevent non-nuclear states from gaining nukes. It had a section about disarmament but that was mostly aspirational.