FR version is available. Content is displayed in original English for accuracy.
Advertisement
Advertisement
⚡ Community Insights
Discussion Sentiment
45% Positive
Analyzed from 1556 words in the discussion.
Trending Topics
#companies#money#going#ram#long#actually#energy#data#more#investors
Discussion (41 Comments)Read Original on HackerNews
The doomer discourse around AI is about as overhyped as the investors are. It's two parties getting pumped up by social media who will be disappointed it's just some new higher economic tier rather than a quick boom or bust.
It always takes twice as long for new markets to fully mature. Industries need to actually adapt technology before they see serious productivity gains (including coding). For now lots of businesses are haphazardly slapping AI on everything and early failures due to immaturity are being used to project long term negative outcomes. While chip and energy development will take even longer as unlike software it's full of the usual baggage of developing things IRL (long term high capital costs, local politics, regulatory compliance, supply chains, etc).
But if it is a bubble, which many would argue is the case, an AI capex bust with it's circular financing collapse is one of, if not THE top threat to global financial stability. ~$725B combined hyperscaler capex in 2026 against roughly $25B of AI service revenue in 2025 on $250B+ of infrastructure spend. Not exactly a promising situation. [3] [4]
This is immediate-picture, I'm not talking about a singularity extinction event. Just what's actually happening right now and the trajectory of the next 4 years.
[1] https://www.publicpower.org/periodical/article/electricity-d...
[2] https://www.iea.org/reports/energy-and-ai/energy-supply-for-...
[3] https://www.tftc.io/bis-annual-report-2026-ai-bubble-circula...
[4] https://alcapitaladvisory.com/research/intelligence/ai-infra...
The long term investments in datacenters and chip manufacturing will likely coincide with with growth in demand from humanoid robots and self-driving cars, so it is likely going to happen anyway over the next decade.
It just will take longer than people expect which creates financial risk, but US capital markets are more resilient than people think.
The SSD cost 30% more than I paid two years ago and was half the capacity (1TB). The RAM cost double what I paid two years ago, was slower, and was half the capacity (32GB).
Now each stick is likely over 240$
2026 has been the year "you're not just X, you're Y" creeped into sermons and birthday cards. I want off this roller coaster.
I think there are so many people in this same situation, not just with disks (and RAM), but all of the components because of bottlenecks of upgrading one without others, that there's going to be this on-going surge in demand starting from corporate and then trickling through decreasingly sized businesses and then finally the consumer market and this could keep prices higher for longer than most people are expecting.
I recently bought some second hand ex-terminal servers to run homelab stuff. I'd have preferred to get one big-ish thing to run it all, but it was more cost effective to get a handful of smaller, lower-spec devices since that's not what people are trying to get their hands on at the moment.
Based on everything I've seen, this is an all-out sprint to whatever the goalpost is - superintelligence, AGI, singularity. Whoever is first will win, everyone else will lose.
I don't know if it's true or not, but it seems to explain the current direction of tech and leads me to believe no CEO is going to allow his company to slow down.
Big tech can finance a lot of AI investments themselves, but they can't keep up the current investments just by themselves, without compromising their non-AI business.
Tech giants with AI interests, hyperscalers, have used "special purpose vehicles" — shell companies — to quietly issue credit to AI companies, and these AI companies have used this line of credit to purchase/lease compute hardware/infrastructure primarily from their creditors, inflating the demand and price of said hardware.
And much of the future infrastructure has yet to be constructed, and the hardware available now will at some point become obsolete or at least decrease in collateral value.
But the credit doesn't actually come from the hyperscalers themselves, as it exceeds their actual cashflow, so it comes from investment banks and/or private investors/lenders... who actually absorb the majority of the risk then?
And the big banks are currently offloading their loans at discounted rates, while simultaneously trading in swaps against hyperscalers?
I presume the reason for the shell companies/private funding shadiness, is to keep debt off-the-record and perhaps avoid regulation / exceed risk tolerance limits.
Just how exposed are the banks and the hyperscalers in all this?
The fact that these companies are either selling these things at multiples of their previous prices, and even then, their P/E ratios are often 10,20,40 shows there's a bidding war for these chips.
It's too much money chasing a fixed amount of product, and the only way to scale the industry is by scaling the entire supply chain, which is a long and expensive process, and certainly isn't fixed by throwing more money at companies.
If existing hardware was sold at the usual margins, all this stuff would cost a tiny fraction of the current price.
This is clearly a precarious position.
There's some winners though, people dumping AI stocks are mostly buying stocks elsewhere, Apple being an obvious winner[0].
[0] https://www.theguardian.com/technology/2026/jul/28/apple-sec...
The KOSPI is almost not worth talking about as any serious signal. It has a circuit breaker drop almost weekly (again followed in no time with an equally high rebound) and basically has come to represent how insanely the South Korean market has become pure gambling (with retail investors absurdly leveraged). Sure , it's hard to image this doesn't lead to some disaster in the long run, but these fluctuations have become par for the course.
Every stick of DDR5 should come with a share, so future dividends go back to the people who financed this insanity.
This will only lead to hyperscalers getting wealthier.