Crypto -> LLM -> Repurpose Crypto Power Contracts to Feed LLM Datacenters -> AI crash -> Repurpose LLM Hardware/Power for Crypto ?
Would be curious if anyone has run the numbers on what LLM hardware capacity (including upcoming) would do to blockchains, if a big chunk were reallocated.
If you're in a "selling stock" part of your life I understand, but if you're in a "buying stock" part of your life it's worth reflecting on the quote until you can shake the "recent price action IS investor psychology" out of your head.
Even when people are only buying with no intent to sell, they'll often complain if prices fall right after they've bought an item, and ask for a rebate for the difference, because they feel they deserve the new lower price. Many retailers are aware of this psychology and do offer some sort of forward-looking price match, both to help buyers overcome that hesitation, and to avoid going through the paperwork of returning and refunding a used item just to sell the same thing again anyway. But the stock market offers no such buyer protections.
It's not a flaw for people to think from the seller's perspective. For you to be a net buyer, someone else has to be a net seller.
Not to mention the unhappiest of all: the people who borrowed money against the value of the assets they owned.
In some situations you can deduct losses in the stock market from the taxes to pay over the next 5 years. Not the same, but not nothing.
Globally, pension funds hold approximately $5–7 trillion directly in the top US tech stocks (the "Magnificent Seven" and adjacent AI infrastructure).
Also in the last couple of years many pension funds have moved money into Private Equity and Private Credit to chase higher returns and they're the backstop for all the off-books AI datacenter buildout debt?
Is roulette wealth creation and destruction?
- A rule of thumb suggested by one study is that every $100 drop in stock market wealth leads, on average, to a $3.20 drop in consumer spending. Under such an assumption, a dotcom-style crash would cut American consumption by about $890bn, or 2.9% of GDP.
You have exactly same assets as before. The businesses of which you are a fractional owner have the same fundamentals.
But other people won't trade other assets for yours at the same rate.
You spent $X to buy RAM chips, expecting that you could produce $Y with it. But you didn't. So you (1) failed to realize the expected value $Y, and (2) misallocated $X, which in hindsight you would have used differently.
Again, that's all learning that future expectations do not match reality.
The decision/action happened earlier, and is separate from the realization. Attributing the material loss to the realization is misplaced.
More important for the wider economy is how much liquidity it destroyed and the mayhem it will do in the bond markets.
But, since the top 10% of American savers (including almost all of US Congress) put most of their savings in stocks they will get a massive bailout. It will make the 2008 Wall Street bailouts look like pocket change. I bet they will say it's for national security or some other lame excuse.
lotta people buying houses and cars with AI seed company money that funds their 400k salary.
AI has a lot of rich people riding on its success, and this time's different for, IMO, two major reasons...
- First, the companies most invested in AI are perfusing it everywhere. Many parts of these big businesses, if not the business as a whole, is invested heavily in the success of LLM-based products. Microsoft is probably the poster child for this, where you can't use practically any of their modern products without copilot or some such being shoved in your face. OpenAI and Anthropic are both companies whose existence is predicated only by a viable LLM-based product. Nvidia and (as of their last-week's announcement) Micron are both now also heavily invested in the success of this technology, though they're also surely not the only companies in the hardware sector to be following this path to some degree.
- Second, the actual individuals behind these companies are the world's richest people, and much of their fortune comes from stock in these companies, and loans taken out against that stock.
They stand to *personally* lose a significant-even-to-them sum of money if the bottom falls out of this thing. If this weren't the case, I'm absolutely certain we wouldn't be seeing as much reporting about how an AI crash would hurt the average household. When smaller crashes happen (even those that affect more average households), it's inevitable, or it's good for the economy in the long-term (it's a correction, after all -- how could that possibly be bad?), or it's the consequence of people's personal choice to invest one way or another, but because the uber-rich were spared, it's *not really that bad.*
This is a disgusting turn in the state of journalism. I don't pretend know what comes next, or how this can be remedied. Crowdsourced news is as prone to manipulation as "traditional" centralized news, so that's not it, and I don't think people have the depth of knowledge to use something purely fact-based (like bellingcat) for every domain of day-to-day life (which is less an effect of the US education system being in continuous decline over decades, and more an effect of the cognitive load it takes to be familiar enough with everything to be able to draw reasonable conclusions about it.