I’m not sure I understand these references. The banks were too big to fail specifically because they were banks involved with the finances of every major industry and government, not simply because their (arguably specious) valuations, or even market caps, had a ton of zeroes on them. What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy that it can’t be allowed to fail? What’s the argument for how such a bailout would result in greater economic outcomes? The banks that got bailed out continued lending and immediately resumed profitable business, how will the AI companies offer value towards such a proposition?
Too big to fail is an oxymorononic statement. Bailing out bad businesses retains those that poorly managed them. Those organizations should of been sold off to remove the bad actors.
AI is currently a sunk cost to the US stock industry that is repeating the bad actor scenario. Not a single AI company is profitable and none of them produce deterministic nor cost effective solutions
Microsoft's statment of using AI to find the most resource intensive applications being ran highlights this. Task Manager does the same thing and does not need a server farm for training. It also uses MB of RAM vs GB.
If manufacturing had the same error rate in production as AI, those plants would of went out of business.
Both industries heavy use legal bribes, donations. Politicians will gladly bail them out to take ℅ of the cut in bribes.
Too big to fails are false claims to retain the bad actors that fund politicians. Bad actors need to fail so the good ones can properly operate.
Too big to fail is also allowing large corporations to skirt copyright laws. You or I seeding TB of copyright content would be thrown in jail.
Too big to fail is rebranding of legalizing corruption.
> Both industries heavy use legal bribes, donations. Politicians will gladly bail them out to take ℅ of the cut in bribes.
Are you saying that it is likely that at some point in the not too distant future, OpenAI and Anthropic will need bailout-size cash infusions from the US Government to continue existence and that the US government will do it and not face severe political consequences?
I just don't think that chain of events is likely. The current administration pays very close attention voter sentiment.
> Are you saying that it is likely that at some point in the not too distant future, OpenAI and Anthropic will need bailout-size cash infusions from the US Government to continue existence and that the US government will do it and not face severe political consequences?
> The current administration pays very close attention voter sentiment.
In which alternative universe? The amount of bribery and self-enrichment is staggering. The treatment of the war is mind boggling. The actual political moves seem to be designed to punish disloyal republicans rather then win more votes.
> What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy that it can’t be allowed to fail?
AI companies are more and more interweaved with the economy because half the world is owning their stocks or has lent them money. Or, they have invested in companies that in turn have invested in AI. It is very similar to the situation before the credit crisis.
Moreover what would a bailout even look like? The banks got loan guarantees from the government essentially.
But like what happens if the government guarantees open ai's loans if the company is structurally unprofitable? Does the government create an operating subsidy?
The government intervened to prevent massive job losses, protect the domestic auto industry, and, in the 1979 case, preserve critical national security manufacturing, as Chrysler produced the M-1 Abrams tank.
Now, I suppose it is possible to imagine that the US government might bail out either or both of OpenAI or Anthropic (whether or not there's an ROI like there was with the Bank Bailout of the 2008 crisis) if the govt. deemed the technology critical to keep a fast pace on (I think we can say without doubt that requirement is satisfied) but, crucially, the government's calculus is that it is better to have these companies compete rather than bring the knowhow in-house.
Inference is cheap, only the training is expensive. Both Bernie and Trump have suggested doing a partial government takeover of the big AI companies to start a sovereign wealth fund.
So it would look like the government taking ownership, letting investors lose their stake, and then operating as inference-only, which would turn a profit
Where have you looked? OpenRouter? Your own experiments? From running various models locally on my MacBook, and paying for the laptop and the electricity to power it, but not the training run, as all I did was install some software that downloaded models from Hugging Face, yes it's cheap. Well, the hardware was several thousand dollars, so not cheap on a personal level, but not unaffordable either.
OpenRouter is the router. We don't care about their financials, the point is that you can buy inference via them for $x/token on a variety of models for a variety of providers. Those are businesses not propped up by SV VC dreams, just hosting plus compute and their costs.
Running a local LLM isn't a mainstream normal thing to do, sure but saying it's "basically 0 of what regular people expect from state of the art LLMs" is lazily dismissing evidence because it contradicts your beliefs. It does work, and it works at a level somewhere above "basically zero" for nerds who are willing and able to set it up for themselves today. The comparison isn't Apache, it's ElasticSearch. It's not Apache cheap and simple, but it's also not Google Spanner expensive.
> The banks that got bailed out continued lending and immediately resumed
Pushing against regulation supposed to prevent this happening again. They also resumed taking a lot of risks, just like before. Their managers got rich while doing the same decisions as before, because other people paid the price.
> What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy
It does not need to be inherently critical and interweaved with the rest of the economy. It has to pay the bribe to the right president. The "military necessity" excuse will be used then.
The argument has already been made. They argued that their business model would collapse if they weren't allowed to train on a bunch of data that wasn't theirs, and nobody stopped them. The argument will be made even more later, as they will argue that too many companies are dependent on their technology, etc.
That's obviously how the AI labs are trying to position themselves. But slop generators are not integral to anything. They most definitely should be left to fail, and if the market so dictates, the hundreds of billions invested should go to zero.
High growth scenario and medium growth scenario (Graph 2). I feel like an idiot asking - aren't we missing some, or at least one, scenario? Is "medium growth" for the next 4 years really the worst people can think of?
Essentially yes? The stock market operates entirely on the assumption that the lines will keep going up. As soon as they flatten the whole thing collapses onto itself.
Flatten would mean there is a profitable business model. It's been years since people have been too tired of repeatedly asking "where will the profit come from?" with no answer. This shit has exactly one direction it will end and it's not flat or up.
The source article says (in a way that I can understand might be a bit non-obvious) that it's analyzing the AI buildout from the perspective of the people who are participating in it. Someone who believes there'll be less than "medium growth" would probably not be buying hyperscaler bonds right now regardless of their precise estimate. They discuss the tail risks if the whole investment thesis is wrong at the end.
Can't you see how much money is being pumped into this lunacy? Of course it's going to succeed. Graphs for failire are such a bummer too and are bad for the economy...
if growth doesn't materialize, then the infrastructure build out plays out exactly like the dot com bubble. the biggest difference this time around is the earnings. if those fall, the rest crubmles.
In the dot-com situation earnings "didn't matter" as long as there was growth. We[1] all believed profit would come eventually. The lesson learned from that experience was that earning do sort of matter. It turns out there is a limit to selling dollars for dimes. Since then revenue, profit and unit-economics ("fundamentals") have gotten almost as much attention as they deserve.
[1] a broad and poorly defined group of "we" - typically investors and tech-bro types.
Earnings did come, and the outcome of the internet include Google, Amazon, and others -- several of the ten biggest companies in the world.
In 2000, there was no sane way to predict who the success stories versus failures would be, timeliness, or otherwise.
AI will be a big change. We don't know how big, when, or who the losers and winners will be yet. Everything could be grossly overvalued or undervalued. We'll only know in hindsight.
Usefulness aside, I see little evidence AI is making money (profit, not revenue) for any firm whose profit doesn't come from the AI itself or the infrastructure, including supply chain. I'd love to hear a counterexample. One such example would be of a hypothetical company that does translations for payment, and with AI they now are making more profit because they use AI to do the translation rather than pay a translator.
Duolingo is such a company you would expect AI to help a lot. Surely AI could allow it to cut costs substantially. And yet, in the past year its stock is down 70% and in Q1 2026 profit has not seemed to increase compared to Q4 2025. In fact, other than Q3 of last year which had some tax shenanigans, their profit is relatively flat. Not a great look given that AI is highly disruptive to their product.
---
AI is actually insidious. Suppose you're in a competitive industry like Costco making 3% (net profit) margin. Suppose the average costco employee makes 60K. Then you come in and think it would be great to have an AI agent lets every employee ask questions of inventory to help customers. Surely if employees could use AI that could somehow make more money for Costco. Hypothetically let's say this ends up costing about the same as the basic subscription in terms of tokens. $20/employee/month Can't be that bad right?
$240 ÷ 0.03 = $8,000 (in other words, generate over 10% of their own salary in marginal additional net profit every year). Is Costco really going to generate 8K more per employee? Nope. And yet, firms like Costco who choose AI effectively just lower their own profit margins.
I want to add context for Duolingo because I think it's confounded by cultural conversation. Duolingo in some Q1 earnings call stated it was going to go AI-native. They were going to switch from contractors to AI. This led to a huge PR crisis, especially on TikTok, where they used to have a big online presence and GenZ influence. I'm not sure if they ever recovered the image they had after that.
They tried, and they got pushback from their consumer base.
Your Costco example makes a lot of sense to me. Right now at my company we are spending hundreds and thousands per employee, but it's not like everyone has an idea that is meant to be integrated into product. So everyone's just making more vaporware.
I do think this is the year the numbers and projections will start coming down to the ground. We can already see this with the fact that the leaders at the frontier labs have stopped talking about AGI and have started talking about token costs and just direct comparisons. It's no longer pie in the sky.
This study finds increased sales and value per customer from GenAI integration at a large Chinese online retailer (all the way back in 2023-24!) The customer Q&A scenario is one of those covered, except the customer talks directly to the LLM rather than an employee with a subscription:
> We quantify the short-term impact of Generative Artificial Intelligence (GenAI) on sales performance through a series of large-scale randomized field experiments involving millions of users and products at a leading cross-border online retail platform. Over 2023-2024, the platform integrated GenAI into seven business workflows spanning customer service, consumer-product matching, advertising, and seller services. We find that GenAI adoption increases sales in most workflows, with effects ranging from no detectable impact to 16.3%, depending on GenAI's marginal contribution relative to baseline firm practices. Across the four GenAI applications with positive sales effects, the implied annual incremental value is roughly $5 per consumer−an economically meaningful impact given the retailer's scale and the early stage of GenAI adoption. The gains operate primarily through higher conversion rates rather than larger cart values, consistent with GenAI improving the shopping experience by reducing search, information, communication, and personalization frictions. Importantly, these effects are not associated with worse post-purchase outcomes, as product return rates and customer ratings do not deteriorate. Finally, we document substantial demand-side heterogeneity, with larger gains for less experienced consumers. Our findings provide novel, large-scale causal evidence on how GenAI shapes sales productivity in online retail, highlighting both its immediate value and broader potential.
Impact on profitability itself is hard to determine due to caveats listed in the paper (which are important to read!) but offhand I would guess that incemental $5 margin per customer is much more than what their prompts cost.
I'm not an MBA over here, but this math seems wrong. If they are spending $240 in increased costs, then they only have to make about $247 in additional revenue from that spend to preserve a 3% margin. That seems much more reasonable if it increases the probability that customers find the product they are looking for and have a good experience.
I agree here. OP is taking a retail company's entire profit margin, which includes a lot of operating costs, and estimating that the AI subscription will have the same margin. The AI subscription is software though, it probably has the operating costs and profit margins of software.
Translator services usually get paid because they offer an accreditation or certification that the humans doing the translation are trained and are not randos / computers. I could imagine these services becoming marginally more efficient but no dramatically improving profits.
The risk here is that either AI commoditizes the software "why ask duolingo to ask chatgpt when I can do so directly?" or simply adds cost to an existing service "why does every costco employee now cost 10% more?"
So far, there hasn't been a clear win for "software wrapping an existing LLM" - if the AI is good enough, then the users can go directly to the source.
I agree with your main observations, but the Costco example is a bit contrived. I expect companies to eventually figure out suitable applications for AI, and I doubt a flat subscription per seat will be one of them. Personally, I feel the main issue is that the tooling + systems needed to deploy AI successfully have only recently started to mature.
AI has been making a ton for advertisers for decades. Think Facebook ads, Google search ads, etc.
That's a different and older kind of AI than chatbots, but it's not fundamentally different, but AI is the engine that makes their ad targeting so effective, and why they're some of the most profitable large companies on the planet.
You’re talking about machine learning, that’s not what all this money has been ploughed into, and not what people mean when they say ‘AI’ today but you’re right that it’s much more clearly effective and profitable than LLMs.
I don't know why you would expect Duolingo profits be helped a lot by AI. Their profits going up and down are more about how much then enshittify for free users - basically tradeoff between long term success and temporary profits. AI does not change that balance. You can enshittify and profit quickly or not enshittify and keep long term engagement.
AI announcement annoyed some people, the slop translate courses were, well, slop. That is the extend of the change.
Speaking of financing: how is the Anthropic IPO going, what is the timeline? They filed over a month ago, no news since. (I would have expected some spectacular news headlines that would be designed to fuel public interest in the impending IPO, but can't detect anything of substance)
just recalled: some significant open source projects are being rewritten in Rust, with the help of Claude. I don't know if these efforts will be successful in the long run, but in some way these news headlines may be creating a media dynamic as part of the IPO preparations?
AFAICT, the last we heard of any AI company IPO was that OpenAI got spooked by the market response to SpaceX and is considering punting to 2027 (https://www.the-independent.com/tech/openai-ipo-date-valuati...). My money is on Anthropic similarly punting, especially if SpaceX manages to cross the very very short distance remaining to drop below the IPO price.
Public utilities tend to pass on all their costs to their customers. If the data centers crumble, that just means the remaining customers (business and residential) each will have to pick up a larger share of the debt payments for the buildout.
> No, we won't - there's significant capex on all that infra that will have to be paid down, and we won't have datacenters to help pay for it.
AIUI, OP is saying that, with all these DCs with no load, we'll have excess electricity generation capacity that was built to support these DCs. That's the "cheap power" he is talking about, not necessarily "cheap computational power".
Thinking out loud, is productivity the ultimate macro benefit of AI? Should we expect macro AI investment to be a leading indicator of macro productivity gains?
For example, did macro investment in factory automation predict future productivity gains?
I've seen other reports that suggest the level of investment for eclipses the internet buid out in 2000 and the railroad boom more than a century earlier. I wonder if they use different ways of landing on these wildly different assessments
Yes. In inflation adjusted dollars spending on AI dwarfs previous "megaprojects". But as a fraction of GDP it's fairly modest -- comparable to the Apollo Project.
It's a sign of how much the economy has grown that under "1% of GDP for a few years" now is far bigger than "over 10% of GDP for a few decades" was in the late 1800s.
You seem to be implying that railway spending was "over 10% of GDP for a few decades" in the late 1800s. If yes then can you trace that back to a methodology? I tried and found much lower numbers, around 3% average over the peak decade.
Your estimate of current AI spending is also low. Hyperscaler capex alone is around 2% of US GDP, not including other costs (neoclouds, employee comp, etc.).
BIS released a larger report in June that identified AI financing/sustainability as one of the biggest risks for the global economy:
https://www.bis.org/publ/arpdf/ar2026e.htm
Thread: https://news.ycombinator.com/item?id=48912577
pre-echos of "too big to fail"
I’m not sure I understand these references. The banks were too big to fail specifically because they were banks involved with the finances of every major industry and government, not simply because their (arguably specious) valuations, or even market caps, had a ton of zeroes on them. What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy that it can’t be allowed to fail? What’s the argument for how such a bailout would result in greater economic outcomes? The banks that got bailed out continued lending and immediately resumed profitable business, how will the AI companies offer value towards such a proposition?
Too big to fail is an oxymorononic statement. Bailing out bad businesses retains those that poorly managed them. Those organizations should of been sold off to remove the bad actors.
AI is currently a sunk cost to the US stock industry that is repeating the bad actor scenario. Not a single AI company is profitable and none of them produce deterministic nor cost effective solutions
Microsoft's statment of using AI to find the most resource intensive applications being ran highlights this. Task Manager does the same thing and does not need a server farm for training. It also uses MB of RAM vs GB.
If manufacturing had the same error rate in production as AI, those plants would of went out of business.
Both industries heavy use legal bribes, donations. Politicians will gladly bail them out to take ℅ of the cut in bribes.
Too big to fails are false claims to retain the bad actors that fund politicians. Bad actors need to fail so the good ones can properly operate.
Too big to fail is also allowing large corporations to skirt copyright laws. You or I seeding TB of copyright content would be thrown in jail.
Too big to fail is rebranding of legalizing corruption.
> Both industries heavy use legal bribes, donations. Politicians will gladly bail them out to take ℅ of the cut in bribes.
Are you saying that it is likely that at some point in the not too distant future, OpenAI and Anthropic will need bailout-size cash infusions from the US Government to continue existence and that the US government will do it and not face severe political consequences?
I just don't think that chain of events is likely. The current administration pays very close attention voter sentiment.
> Are you saying that it is likely that at some point in the not too distant future, OpenAI and Anthropic will need bailout-size cash infusions from the US Government to continue existence and that the US government will do it and not face severe political consequences?
Didn't they say this themselves? https://edition.cnn.com/2025/11/06/tech/openai-backtracks-go...
> The current administration pays very close attention voter sentiment.
Is that how the Anti-Weaponization Fund came about?
It wouldn’t be the first time the US government made expensive and unpopular moves in the name of national security and walked away largely unscathed.
> The current administration pays very close attention voter sentiment.
In which alternative universe? The amount of bribery and self-enrichment is staggering. The treatment of the war is mind boggling. The actual political moves seem to be designed to punish disloyal republicans rather then win more votes.
> What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy that it can’t be allowed to fail?
AI companies are more and more interweaved with the economy because half the world is owning their stocks or has lent them money. Or, they have invested in companies that in turn have invested in AI. It is very similar to the situation before the credit crisis.
Moreover what would a bailout even look like? The banks got loan guarantees from the government essentially.
But like what happens if the government guarantees open ai's loans if the company is structurally unprofitable? Does the government create an operating subsidy?
Modeled on Chrysler, perhaps?
> Modeled on Chrysler, perhaps?
The government intervened to prevent massive job losses, protect the domestic auto industry, and, in the 1979 case, preserve critical national security manufacturing, as Chrysler produced the M-1 Abrams tank.
Now, I suppose it is possible to imagine that the US government might bail out either or both of OpenAI or Anthropic (whether or not there's an ROI like there was with the Bank Bailout of the 2008 crisis) if the govt. deemed the technology critical to keep a fast pace on (I think we can say without doubt that requirement is satisfied) but, crucially, the government's calculus is that it is better to have these companies compete rather than bring the knowhow in-house.
Is that what you're thinking?
Inference is cheap, only the training is expensive. Both Bernie and Trump have suggested doing a partial government takeover of the big AI companies to start a sovereign wealth fund.
So it would look like the government taking ownership, letting investors lose their stake, and then operating as inference-only, which would turn a profit
Inference is cheap as long as you have 1000 data centers full of GPUs. The only crux is you have neither.
> Inference is cheap
We have absolutely 0 hard proof of this. We have a lot of wishful thinking but no hard numbers, audited numbers from any public entity.
I'd love to see them if they are available.
Where have you looked? OpenRouter? Your own experiments? From running various models locally on my MacBook, and paying for the laptop and the electricity to power it, but not the training run, as all I did was install some software that downloaded models from Hugging Face, yes it's cheap. Well, the hardware was several thousand dollars, so not cheap on a personal level, but not unaffordable either.
> OpenRouter
Do we have the balance sheet for OpenRouter & co?
Especially in this age where if you put AI in your company's mission statement you're drowned in money.
Let's hold off on calling something "cheap" until the external financing money runs out and the actual numbers are revealed AND audited.
> yes it's cheap.
When running toy models that do basically 0 of what regular people expect from state of the art LLMs, sure.
Running Apache is cheap. Running Google search isn't. They both serve web pages.
OpenRouter is the router. We don't care about their financials, the point is that you can buy inference via them for $x/token on a variety of models for a variety of providers. Those are businesses not propped up by SV VC dreams, just hosting plus compute and their costs.
Running a local LLM isn't a mainstream normal thing to do, sure but saying it's "basically 0 of what regular people expect from state of the art LLMs" is lazily dismissing evidence because it contradicts your beliefs. It does work, and it works at a level somewhere above "basically zero" for nerds who are willing and able to set it up for themselves today. The comparison isn't Apache, it's ElasticSearch. It's not Apache cheap and simple, but it's also not Google Spanner expensive.
Except for movie pass startups are marginally profitable on whatever they sell.
But you have to keep training
The value of AI and related companies on retirement funds
> The banks that got bailed out continued lending and immediately resumed
Pushing against regulation supposed to prevent this happening again. They also resumed taking a lot of risks, just like before. Their managers got rich while doing the same decisions as before, because other people paid the price.
> What’s the argument for OpenAI being so inherently critical and interweaved with the rest of the economy
It does not need to be inherently critical and interweaved with the rest of the economy. It has to pay the bribe to the right president. The "military necessity" excuse will be used then.
Who do you think the private credit lenders are exactly?
The argument has already been made. They argued that their business model would collapse if they weren't allowed to train on a bunch of data that wasn't theirs, and nobody stopped them. The argument will be made even more later, as they will argue that too many companies are dependent on their technology, etc.
“National security! We can’t let China win the AI race!” Or some BS
That's obviously how the AI labs are trying to position themselves. But slop generators are not integral to anything. They most definitely should be left to fail, and if the market so dictates, the hundreds of billions invested should go to zero.
The thing about bailing OpenAI or Anthropic is that they would need a new bailout every few months.
High growth scenario and medium growth scenario (Graph 2). I feel like an idiot asking - aren't we missing some, or at least one, scenario? Is "medium growth" for the next 4 years really the worst people can think of?
Financial news tends to be written for people who can fill in a lot of blanks themselves.
Can't agree more, tech and finance bros have a lot in similar except when it comes to business.
Essentially yes? The stock market operates entirely on the assumption that the lines will keep going up. As soon as they flatten the whole thing collapses onto itself.
Flatten would mean there is a profitable business model. It's been years since people have been too tired of repeatedly asking "where will the profit come from?" with no answer. This shit has exactly one direction it will end and it's not flat or up.
Flatten plus debt is bad.
Debt presumes future growth.
> Is "medium growth" for the next 4 years really the worst people can think of?
At this point anything less than "medium growth" will crash the economy. We'll have bigger problems if that happens (think 2000 or 2008)
Right... So since it's a big problem, shouldn't we at least be considering it as a possibility so that we can minimize the impact?
No no no… that’s unecessary. Because it won’t happen.
Hmm. Perhaps too similar to pre-GFC when the ratings agencies' models never accounted for scenarios where home prices went down at the national level.
The source article says (in a way that I can understand might be a bit non-obvious) that it's analyzing the AI buildout from the perspective of the people who are participating in it. Someone who believes there'll be less than "medium growth" would probably not be buying hyperscaler bonds right now regardless of their precise estimate. They discuss the tail risks if the whole investment thesis is wrong at the end.
Can't you see how much money is being pumped into this lunacy? Of course it's going to succeed. Graphs for failire are such a bummer too and are bad for the economy...
if growth doesn't materialize, then the infrastructure build out plays out exactly like the dot com bubble. the biggest difference this time around is the earnings. if those fall, the rest crubmles.
How are the earnings different this time? Can you add any color to that?
In the dot-com situation earnings "didn't matter" as long as there was growth. We[1] all believed profit would come eventually. The lesson learned from that experience was that earning do sort of matter. It turns out there is a limit to selling dollars for dimes. Since then revenue, profit and unit-economics ("fundamentals") have gotten almost as much attention as they deserve.
[1] a broad and poorly defined group of "we" - typically investors and tech-bro types.
That's not quite the lesson.
Earnings did come, and the outcome of the internet include Google, Amazon, and others -- several of the ten biggest companies in the world.
In 2000, there was no sane way to predict who the success stories versus failures would be, timeliness, or otherwise.
AI will be a big change. We don't know how big, when, or who the losers and winners will be yet. Everything could be grossly overvalued or undervalued. We'll only know in hindsight.
Usefulness aside, I see little evidence AI is making money (profit, not revenue) for any firm whose profit doesn't come from the AI itself or the infrastructure, including supply chain. I'd love to hear a counterexample. One such example would be of a hypothetical company that does translations for payment, and with AI they now are making more profit because they use AI to do the translation rather than pay a translator.
Duolingo is such a company you would expect AI to help a lot. Surely AI could allow it to cut costs substantially. And yet, in the past year its stock is down 70% and in Q1 2026 profit has not seemed to increase compared to Q4 2025. In fact, other than Q3 of last year which had some tax shenanigans, their profit is relatively flat. Not a great look given that AI is highly disruptive to their product.
---
AI is actually insidious. Suppose you're in a competitive industry like Costco making 3% (net profit) margin. Suppose the average costco employee makes 60K. Then you come in and think it would be great to have an AI agent lets every employee ask questions of inventory to help customers. Surely if employees could use AI that could somehow make more money for Costco. Hypothetically let's say this ends up costing about the same as the basic subscription in terms of tokens. $20/employee/month Can't be that bad right?
$240 ÷ 0.03 = $8,000 (in other words, generate over 10% of their own salary in marginal additional net profit every year). Is Costco really going to generate 8K more per employee? Nope. And yet, firms like Costco who choose AI effectively just lower their own profit margins.
I think this is a great point.
I want to add context for Duolingo because I think it's confounded by cultural conversation. Duolingo in some Q1 earnings call stated it was going to go AI-native. They were going to switch from contractors to AI. This led to a huge PR crisis, especially on TikTok, where they used to have a big online presence and GenZ influence. I'm not sure if they ever recovered the image they had after that.
They tried, and they got pushback from their consumer base.
Your Costco example makes a lot of sense to me. Right now at my company we are spending hundreds and thousands per employee, but it's not like everyone has an idea that is meant to be integrated into product. So everyone's just making more vaporware.
I do think this is the year the numbers and projections will start coming down to the ground. We can already see this with the fact that the leaders at the frontier labs have stopped talking about AGI and have started talking about token costs and just direct comparisons. It's no longer pie in the sky.
The issue is Duolingo is rubbish, what the app does isn’t that hard to replicate and there are already competitors that have a better product.
So if AI is real then that‘s the cherry on top: people can now make an alternative to your ineffective messy app even easier.
For those kind of SaaS products with no moat LLMs could actually be a problem and definitely aren’t a good thing
This study finds increased sales and value per customer from GenAI integration at a large Chinese online retailer (all the way back in 2023-24!) The customer Q&A scenario is one of those covered, except the customer talks directly to the LLM rather than an employee with a subscription:
https://arxiv.org/abs/2510.12049
> We quantify the short-term impact of Generative Artificial Intelligence (GenAI) on sales performance through a series of large-scale randomized field experiments involving millions of users and products at a leading cross-border online retail platform. Over 2023-2024, the platform integrated GenAI into seven business workflows spanning customer service, consumer-product matching, advertising, and seller services. We find that GenAI adoption increases sales in most workflows, with effects ranging from no detectable impact to 16.3%, depending on GenAI's marginal contribution relative to baseline firm practices. Across the four GenAI applications with positive sales effects, the implied annual incremental value is roughly $5 per consumer−an economically meaningful impact given the retailer's scale and the early stage of GenAI adoption. The gains operate primarily through higher conversion rates rather than larger cart values, consistent with GenAI improving the shopping experience by reducing search, information, communication, and personalization frictions. Importantly, these effects are not associated with worse post-purchase outcomes, as product return rates and customer ratings do not deteriorate. Finally, we document substantial demand-side heterogeneity, with larger gains for less experienced consumers. Our findings provide novel, large-scale causal evidence on how GenAI shapes sales productivity in online retail, highlighting both its immediate value and broader potential.
Impact on profitability itself is hard to determine due to caveats listed in the paper (which are important to read!) but offhand I would guess that incemental $5 margin per customer is much more than what their prompts cost.
I'm not an MBA over here, but this math seems wrong. If they are spending $240 in increased costs, then they only have to make about $247 in additional revenue from that spend to preserve a 3% margin. That seems much more reasonable if it increases the probability that customers find the product they are looking for and have a good experience.
I agree here. OP is taking a retail company's entire profit margin, which includes a lot of operating costs, and estimating that the AI subscription will have the same margin. The AI subscription is software though, it probably has the operating costs and profit margins of software.
Translator services usually get paid because they offer an accreditation or certification that the humans doing the translation are trained and are not randos / computers. I could imagine these services becoming marginally more efficient but no dramatically improving profits.
The risk here is that either AI commoditizes the software "why ask duolingo to ask chatgpt when I can do so directly?" or simply adds cost to an existing service "why does every costco employee now cost 10% more?"
So far, there hasn't been a clear win for "software wrapping an existing LLM" - if the AI is good enough, then the users can go directly to the source.
I agree with your main observations, but the Costco example is a bit contrived. I expect companies to eventually figure out suitable applications for AI, and I doubt a flat subscription per seat will be one of them. Personally, I feel the main issue is that the tooling + systems needed to deploy AI successfully have only recently started to mature.
AI has been making a ton for advertisers for decades. Think Facebook ads, Google search ads, etc.
That's a different and older kind of AI than chatbots, but it's not fundamentally different, but AI is the engine that makes their ad targeting so effective, and why they're some of the most profitable large companies on the planet.
You’re talking about machine learning, that’s not what all this money has been ploughed into, and not what people mean when they say ‘AI’ today but you’re right that it’s much more clearly effective and profitable than LLMs.
> but it's not fundamentally different
It is. We are talking about LLMs here.
I don't know why you would expect Duolingo profits be helped a lot by AI. Their profits going up and down are more about how much then enshittify for free users - basically tradeoff between long term success and temporary profits. AI does not change that balance. You can enshittify and profit quickly or not enshittify and keep long term engagement.
AI announcement annoyed some people, the slop translate courses were, well, slop. That is the extend of the change.
Speaking of financing: how is the Anthropic IPO going, what is the timeline? They filed over a month ago, no news since. (I would have expected some spectacular news headlines that would be designed to fuel public interest in the impending IPO, but can't detect anything of substance)
just recalled: some significant open source projects are being rewritten in Rust, with the help of Claude. I don't know if these efforts will be successful in the long run, but in some way these news headlines may be creating a media dynamic as part of the IPO preparations?
[1] https://news.ycombinator.com/item?id=48870966 pgrust passes 100% of the Postgres regression tests
[2] https://news.ycombinator.com/item?id=48837877 Rewriting Bun in Rust
[3] https://news.ycombinator.com/item?id=48789325 My AI-built PHP engine in Rust passes 17% of PHP-src tests, renders WordPress (ekinertac.com)
they're probably waiting to see how market reacts to fable5 and gpt 5.6
AFAICT, the last we heard of any AI company IPO was that OpenAI got spooked by the market response to SpaceX and is considering punting to 2027 (https://www.the-independent.com/tech/openai-ipo-date-valuati...). My money is on Anthropic similarly punting, especially if SpaceX manages to cross the very very short distance remaining to drop below the IPO price.
They think things will be better for them in 2027?
This is the pattern I’d expect for their IPOs too, give their current fantasy valuations.
At least if the datacenters usage crashes, we'll have cheap power from all the infra that got built.
Public utilities tend to pass on all their costs to their customers. If the data centers crumble, that just means the remaining customers (business and residential) each will have to pick up a larger share of the debt payments for the buildout.
No, we won't - there's significant capex on all that infra that will have to be paid down, and we won't have datacenters to help pay for it.
> No, we won't - there's significant capex on all that infra that will have to be paid down, and we won't have datacenters to help pay for it.
AIUI, OP is saying that, with all these DCs with no load, we'll have excess electricity generation capacity that was built to support these DCs. That's the "cheap power" he is talking about, not necessarily "cheap computational power".
Call me overly cynical, but I’m willing to bet we’re already footing the bill in more than a few ways.
+GPUs have significantly shorter useful life spans than say, all the dark fiber laid in the 90s
(January 2026)
Thinking out loud, is productivity the ultimate macro benefit of AI? Should we expect macro AI investment to be a leading indicator of macro productivity gains?
For example, did macro investment in factory automation predict future productivity gains?
I'm not sure the current administration is fully driven by macroeconomic arguments.
I've seen other reports that suggest the level of investment for eclipses the internet buid out in 2000 and the railroad boom more than a century earlier. I wonder if they use different ways of landing on these wildly different assessments
Yes. In inflation adjusted dollars spending on AI dwarfs previous "megaprojects". But as a fraction of GDP it's fairly modest -- comparable to the Apollo Project.
It's a sign of how much the economy has grown that under "1% of GDP for a few years" now is far bigger than "over 10% of GDP for a few decades" was in the late 1800s.
You seem to be implying that railway spending was "over 10% of GDP for a few decades" in the late 1800s. If yes then can you trace that back to a methodology? I tried and found much lower numbers, around 3% average over the peak decade.
https://news.ycombinator.com/item?id=44805979
Your estimate of current AI spending is also low. Hyperscaler capex alone is around 2% of US GDP, not including other costs (neoclouds, employee comp, etc.).
You're right, I was remembering the peak and duration of the railway boom but of course it wasn't at the peak for the entire duration.
Manhattan Project: $36B, 5 years
▫ Apollo Program: $257B, 14 years
▫ Interstate Highway System: $620B, 37 years
▫ AI data centers: $930B, 6 years and still accelerating
From: https://substack.com/@rubendominguez/note/c-244929068
That AI number is a gross underestimate. It’s nearly that for this year alone. Wildly off.
this time its free printed fiat debt tho, not fully comparable. If market crashes, they will print even few times more again