In almost all scenarios, a setup with this incentive structure will lead to massive adoption. It's too tempting, and with most jobs / political positions being short term (<5yrs) ones, people optimize for their time in that timeframe, not longer.
Boards will pursue stock buybacks (short term growth, long term may cause trouble if there's a downturn), banks will lend out subprime mortgages (hit your sales numbers in the short term, at the cost of long term risk), etc etc.
This situation is no different. There's money flowing in and there's less red tape since everyone is being pressured to allow it. It might work out in the long term, it might not, but it will 100% benefit those who push it in the short term. People will get promoted for driving a new data center, politicians can promote more jobs being added, everybody wins... for now.
The future economic aspect becomes irrelevant when the short term candy is sweet enough.
Meanwhile in Australia we have a "AI data centre" startup being valued at $1.9 billion and given $330 million to play with, having not built anything yet. It's co-founded by a guy that went to prison for insider trading. His wife is also an investor, who happens to be a prominent Australian influencer. The company previously focused on Bitcoin mining but have pivoted to the AI boom, claiming their cooling systems to be 60% better than competitors. Their first project will kick off soon in the Australian state of Tasmania, where there is nothing much more than sheep and tourists.
Regardless of a bubble or not, don't some operators do exceptionally well financially, without a lot of recognition, when things are murky compared to transparent or having clarity?
And how long has this kind of thing probably been going on?
Unless the AI industry can figure out how to be profitable soon (to which basically nobody has a clear path to profitability besides maybe ad revenue) it's hard to see this not blowing up in a year or two. The bills for all of this are going to come due eventually and the AI CEOs can only convince investors to keep letting them burn billions for so long.
>> Cheerleaders such as Sam Altman, OpenAI’s boss, argue that the risks of underbuilding are at least as serious as those of overbuilding, because of the long-term economic potential of generative AI.
Oracle Sinks on Report Its Cloud Margins Are Lower Than Expected
Jeran Wittenstein
October 7, 2025 at 6:06 PM GMT+2
Oracle Corp. shares tumbled after a report that the software maker’s profit margin in its cloud computing business is lower than many on Wall Street have been estimating.
While Oracle generated roughly $900 million in revenue from the rental of servers powered by Nvidia Corp. chips during the three months ended in August, the company only managed about $125 million in gross profit, the Information reported, citing internal corporate documents.
Also, this was remarkably hard to find. Might I suggest if the logged in user is looking over their own comments, eg me viewing https://news.ycombinator.com/threads?id=fragmede, that a flagged dead comment with a moderator responding shouldn't be collapsed by default? If that's intentional then nevermind, but as I am truly grateful for this site and you and DanG's (and everyone else's) hard work, I want to point out that it was hard to find and wouldn't have come across it if I didn't have a program running that hits the Firebase API and notifies me on replies. Someone who didn't have such a program could easily miss mod responses.
Cloud has failed (by becoming more expensive then on prem) so they're buying up everything in a last ditch effort to centralize compute. Really silly since all those GPUs will be obsolete soon.
Last weekend we had dinner at a friend to celebrate their child's birthday. They work at OCI in a very senior position and were heading out next day to troubleshoot that exact datacenter in Abilene, TX next day, to presumably yell at various people about cabling and schedule and whatnot. They were not looking forward to it because a) TX and b) it's Abilene TX and c) the overall situation.
We spoke quite a bit about this project, circular money and things and the impression I got is that OCI is just grabbing money while the perpetuum mobile is unrolling.
I am looking forward to hearing how the trip went when I see them next.
Every time I read one of these articles the main issue I have is that it doesn't take into account the huge shortages of compute that are going on all the time. Anthropic and Google especially have been incredibly unreliable, struggling to keep up with demand.
Each of the main providers could easily use 10x the compute tomorrow (albeit arguably inefficiently) by using more thinking for certain tasks.
Now - does that scale to the 10s of GWs of deals OpenAI is doing? Probably not right now, but the bigger issue as the article does point out in fairness is the huge backlog of power availability worldwide.
Finally, AI adoption outside of software engineering is incredibly limited at work. This is going to rapidly change. Even the Excel agent Microsoft has recently launched has the potential to result in hundred fold increases in token consumption per user. I'm also suspect of the AI sell through rate being an indicator that it's not popular for Microsoft. The later versions of M365 copilot (or whatever it is called today) are wildly better than the original ones.
It all sort of reminds me of Apple's goal of getting 1% in cell phone market share, which seemed laughably ambitious at one point - a total stretch goal. Now they are up to 20% and smartphone penetration as a whole is probably close to 90% globally of those that have a phone.
One potential wild card though for the whole market is someone figuring out a very efficient ASIC for inference (maybe with 1.58bit). GPUs are mostly overkill for inference and I would not be surprised if 10-100x efficiency gains could be had on very specialised chips.
> Each of the main providers could easily use 10x the compute tomorrow (albeit arguably inefficiently) by using more thinking for certain tasks.
That's true for everyone with regard to any resource.
The question is whether the 10x increase in resources results in 10x or more increase in profit.
If it doesn't then it doesn't make sense to pay for the extra resources. For AI right now, the constraint is profit per resource unit, not number or resource units.
the huge demand exists right now because the cost of a token is near zero. and companies have figured out one weird hack to gaining value in the stock market, which is to brag about how many tokens are being crammed into all manner of places that they may or may not belong.
customer value must eventually flow out of those datacenters in the opposite direction to the the energy and capex that are flowing in
do people actually want all this AI? I see studio ghibli portraits, huge amounts of internet spam, workslop... where is the value?
"The later versions of M365 copilot (or whatever it is called today) are wildly better than the original ones."
I find AI agents work very poorly within the Microsoft ecosystem. They can generate great HTML documents (because it's an open source format maybe?) but for word documents, the formatting is so poor I'd had to turn it off and just do things manually.
Opposing anecdote: I got consistent performance out of Grok and Qwen (17 providers on Openrouter) throughout the day but Gemini gets slow and dumb at times.
* questionable demand for AI products make the DC investments risky (makes sense)
* DCs being built in “remote” areas with cheap land may become obsolete/replaced by other DCs making use of said cheap land (questionable; DCs can upgrade cheaper than building a new one)
* financing for DCs used to be from Big Tech but is now spread out among private equity, sovereign wealth funds etc increasing the exposure of the economy to failure of these investments (again, questionable, unless they are being financed by bank loans)
The most salient concern seems to be a lack of demand. I don’t see why that would change in the future.
28 comments
[ 3.5 ms ] story [ 58.4 ms ] threadIn almost all scenarios, a setup with this incentive structure will lead to massive adoption. It's too tempting, and with most jobs / political positions being short term (<5yrs) ones, people optimize for their time in that timeframe, not longer.
Boards will pursue stock buybacks (short term growth, long term may cause trouble if there's a downturn), banks will lend out subprime mortgages (hit your sales numbers in the short term, at the cost of long term risk), etc etc.
This situation is no different. There's money flowing in and there's less red tape since everyone is being pressured to allow it. It might work out in the long term, it might not, but it will 100% benefit those who push it in the short term. People will get promoted for driving a new data center, politicians can promote more jobs being added, everybody wins... for now.
The future economic aspect becomes irrelevant when the short term candy is sweet enough.
All of the big players - Nvidia, OpenAI, Oracle, Microsoft - are in insane circular financing agreements that would make Enron executives blush.
https://www.smartcompany.com.au/startupsmart/firmus-raises-3...
And how long has this kind of thing probably been going on?
Pure hucksterism.
Oracle Sinks on Report Its Cloud Margins Are Lower Than Expected
Jeran Wittenstein October 7, 2025 at 6:06 PM GMT+2
Oracle Corp. shares tumbled after a report that the software maker’s profit margin in its cloud computing business is lower than many on Wall Street have been estimating.
While Oracle generated roughly $900 million in revenue from the rental of servers powered by Nvidia Corp. chips during the three months ended in August, the company only managed about $125 million in gross profit, the Information reported, citing internal corporate documents.
We detached this comment from https://news.ycombinator.com/item?id=45510207 and marked it off topic.
I started with mainframes and I'm not going back.
We spoke quite a bit about this project, circular money and things and the impression I got is that OCI is just grabbing money while the perpetuum mobile is unrolling.
I am looking forward to hearing how the trip went when I see them next.
Each of the main providers could easily use 10x the compute tomorrow (albeit arguably inefficiently) by using more thinking for certain tasks.
Now - does that scale to the 10s of GWs of deals OpenAI is doing? Probably not right now, but the bigger issue as the article does point out in fairness is the huge backlog of power availability worldwide.
Finally, AI adoption outside of software engineering is incredibly limited at work. This is going to rapidly change. Even the Excel agent Microsoft has recently launched has the potential to result in hundred fold increases in token consumption per user. I'm also suspect of the AI sell through rate being an indicator that it's not popular for Microsoft. The later versions of M365 copilot (or whatever it is called today) are wildly better than the original ones.
It all sort of reminds me of Apple's goal of getting 1% in cell phone market share, which seemed laughably ambitious at one point - a total stretch goal. Now they are up to 20% and smartphone penetration as a whole is probably close to 90% globally of those that have a phone.
One potential wild card though for the whole market is someone figuring out a very efficient ASIC for inference (maybe with 1.58bit). GPUs are mostly overkill for inference and I would not be surprised if 10-100x efficiency gains could be had on very specialised chips.
That's true for everyone with regard to any resource.
The question is whether the 10x increase in resources results in 10x or more increase in profit.
If it doesn't then it doesn't make sense to pay for the extra resources. For AI right now, the constraint is profit per resource unit, not number or resource units.
customer value must eventually flow out of those datacenters in the opposite direction to the the energy and capex that are flowing in
do people actually want all this AI? I see studio ghibli portraits, huge amounts of internet spam, workslop... where is the value?
I find AI agents work very poorly within the Microsoft ecosystem. They can generate great HTML documents (because it's an open source format maybe?) but for word documents, the formatting is so poor I'd had to turn it off and just do things manually.
* questionable demand for AI products make the DC investments risky (makes sense)
* DCs being built in “remote” areas with cheap land may become obsolete/replaced by other DCs making use of said cheap land (questionable; DCs can upgrade cheaper than building a new one)
* financing for DCs used to be from Big Tech but is now spread out among private equity, sovereign wealth funds etc increasing the exposure of the economy to failure of these investments (again, questionable, unless they are being financed by bank loans)
The most salient concern seems to be a lack of demand. I don’t see why that would change in the future.