For anyone old enough to remember attending bankruptcy/liquidation auctions of "dotcom 1.0" companies in 2000, 2001 or so and buying an Aeron chair, or similar, I really wonder what it will look like this time around. There was one point in time where only very slightly used datacenter cooling systems and diesel backup generators were selling for pennies on the dollar.

▲

Gareth321 8 hours ago | parent | next [-]

IMHO I think the blast radius is implicitly contained. The major investments are in data centres, and the current investment cases impute near zero residual value after five years. Meaning that current valuations already assume "catastrophic" declines in equipment valuation. This is unusually clear-eyed and sober investment calculus in the tech space.

Further, unlike during the dotcom crisis, most of this spending is not driven by debt. It's mostly funded by the large companies which are producing enormous revenue and profit to pay for this. It's an order of magnitude different.

The major question mark on these valuations is the revenue assumptions, which can be reasonably criticised. A bear case here is that revenue growth is not as aggressive as projected, and valuations steadily decline over time. I don't see a likely scenario where the entire sector collapses. There is no apparent cascade failure mechanism. Of course, these mechanisms aren't always immediately clear prior to crashes.

We must remember that all of the models coming out of China are presumed to be distilled frontier models. Meaning a) they will always be x days/weeks/months behind the frontier models, b) they will never be quite as good, c) inference will generally be constrained by compute capacity (especially as the frontier studios have an incentive to capitalise on their moat), meaning Chinese studios will always be at a disadvantage.

The real wildcard here is self-improvements. It looks like we're already in the singularity, meaning a large proportion of LLM development is already done by LLMs. The development cycle on these might be months now, but it will be weeks soon. Days within a year, then hours, minutes, seconds, and milliseconds. It's impossible to predict what this curve looks like.

▲

walrus01 7 hours ago | parent | next [-]

> I don't see a likely scenario where the entire sector collapses. There is no apparent cascade failure mechanism. Of course, these mechanisms aren't always immediately clear prior to crashes.

I can see a scenario where companies like openAI and Anthropic do go belly up but the technology and IP and physical assets remain, get balkanized or snapped up by various other parties. Let's say for instance that they do finish the physical construction of the "Stargate" datacenters in Texas, and they get filled with the equivalent of a whole shitton of B300 RAM/GPU systems and are up and operational. Those don't disappear.

In some kind of catastrophic failure scenario it could end up as a debtor in possession arrangement, or chapter 7 sale to new set of people who want to make use of it. Not unlike what happened to a number of much smaller scale datacenters that were built with 1996-2000 dotcom 1.0 boom money that changed ownership around the 2001-2002 time frame.

	▲	Gareth321 7 hours ago \| parent [-]
		Even if OpenAI or Anthropic or both fail, they're not even listed yet. It barely makes a dent on the stock market. Their infrastructure and IP is purchased by Microsoft/Google/Meta, and those companies valuations adjust accordingly.

▲

ItsBob 7 hours ago | parent | prev [-]

> Further, unlike during the dotcom crisis, most of this spending is not driven by debt.

Unfortunately, that's not the case. Between the big 5 (Microsoft, Meta, Amazon etc.) they're spending more than $600 Billion in 2026! They don't have that much cash lying around so they're selling bonds!

That's debt!

Not only that, they're increasing the bond sales in Europe! I assume that means they're tapped-out in the US!

All this off-the-books stuff, despite being legal but shady, is still debt! Debt has to be paid by someone.

To sum it up: the AI buildout is a highly leveraged, debt-fueled expansion, not an organic, cash-funded software cycle... this will not end well!

▲

Gareth321 7 hours ago | parent [-]

Bond sales are pretty normal for companies, AI or not. Leverage is in very low ratios compared to the dotcom era.

Much of the dotcom era telecom expansion depended on debt and projected future demand, but the resulting networks were so overbuilt that only about 2% of North American long-distance capacity was being used. Falling prices left numerous operators unable to service their debts. Today’s AI expansion is led mainly by already-profitable companies with large existing revenues and cash flows. Microsoft alone generated $136B in operating cash flow during its 2025 financial year while spending $65B on capex. It held $95B in cash, compared with total debt of about $43B. That means a disappointing return on AI investment would reduce profits, cash reserves, and shareholder value, rather than making these companies insolvent.

▲

ItsBob 6 hours ago | parent [-]

I don't think Microsoft would fail if it all crashed. Nor would Google or Amazon as they all have alternate revenue streams.

It's the pure AI companies like OpenAI that will hit the wall.

However, the tech companies are increasing their debt enormously. That's the issue.

	▲	Gareth321 4 hours ago \| parent [-]
		OpenAI and Anthropic aren't even listed. Their failures would barely dent the stock market.

▲

ItsBob 8 hours ago | parent | prev [-]

You'll be able to buy used H100s. One careful owner. Never overclocked etc.

They'll be knackered though, unlike a good Herman Miller chair. :-)

We might even be able to afford RAM again!

	▲	walrus01 8 hours ago \| parent [-]
		Unlike 2001, I'm also trying to imagine amateurs (or very small companies) trying to repurpose pieces of whole-rack liquid cooling systems, with varying degrees of possible success or catastrophe. But I'll be holding out for the gently used one owner B300.