[jgbuddy]

You are making the assumption that the models are only used / paid for by 2.5% of the population (your knowledge workers value). There will be new value created by these models which people are happy to pay for which simply did not exist at all before. It is also naive to say that the hyperscalers are going to be expecting a return on this in 5 years, it will be entirely propped up by investments / IPOs as has been the case with any tech company for decades now to reach scale. The hyperscalers are currently spending ~650b combined annually, which they have the cash for and can sell in future compute instantly.

[specproc]

I'm sorry, what the feck does "value creation" mean here? I live in a place where people are so, insanely squeezed from every angle. Wages are stagnant, prices rocketing. Where is the money to pay for this value going to come from?

No one I know feels richer than they did a decade back. I've not been able to meaningfully put up my prices for a decade. People are tired and stressed and scared, particularly scared of a technology everyone keeps telling them will make them redundant.

There is no rising tide lifting all boats, just most of us drowning whilst a few whizz past in their yachts.

I honestly hope these guys faceplant ASAP. Couldn't happen to a nicer bunch of people.

[dirck-norman]

Feelings aren’t fact. A lot of data shows the doomerism is not reflected in the actual numbers and much of it has to do with rapid inflation and continued vibes.

Consumption has risen, inflation adjusted wages have risen for blue collar and white collar alike. Most social mobility has been the middle class moving into the upper middle class, not moving to the lower class.

The main thing holding people back is the housing crisis. This is orthogonal to the value creation of businesses.

Value creation is growth. If it didn’t exist the S&P would still be 42.55$.

[everforward]

> The main thing holding people back is the housing crisis. This is orthogonal to the value creation of businesses.

This feels wholly at odds with saying most social mobility is upwards. So most of the social movement is into a class where a home and vacations are a given, but we also have a growing class of people who can't afford a home? Per BLS, average real wages are down 0.3% YoY https://www.bls.gov/news.release/realer.nr0.htm .

> Value creation is growth. If it didn’t exist the S&P would still be 42.55$.

This reductively assumes "value creation" is the only effect on the S&P pricing. You'll note a ton of graphs correlate with it, e.g. https://tradingeconomics.com/united-states/inflation-cpi is the US inflation rate, which also tracks the S&P pricing. Ie if a company is worth $100 a year ago and inflation was 4%, I'd expect to pay $104 for their stock with 0 value creation whatsoever.

[dirck-norman]

Home ownership is not the definition of economic class.

The s&p has vastly outstripped inflation, this isn’t even an argument. It’s a very bizarre and uninformed opinion to say “inflation is correlated with s&p value”.

In economics, you deduct the inflation rate from growth to get the real rate of return.

I wonder why so many people with such little understanding of financial markets make comments like these.

https://www.multpl.com/inflation-adjusted-s-p-500

[everforward]

> Home ownership is not the definition of economic class.

It was at one point, so if you're saying it doesn't now then your "social movement" is really just goalpost moving. I'm sure the upper class does grow if you just declare that the boundaries of it are now lower. We're all millionaires if we just redefine millionaire to mean "not currently overdrafted".

> In economics, you deduct the inflation rate from growth to get the real rate of return.

Which neglects the impact of inflation on the principal. If I put $100k into the S&P in 2016 I'd have 303k now (303% value). Cumulative inflation during that period was 38.8%, so your metric says I'm at 164% rate of return. $303k today is only worth $218k in 2016 dollars though, so I'm really only up 118% in absolute value.

That's also only using the CPI, and neglecting asset inflation which impacts stock prices outside of actual value creation. More money floating around means more gets parked in index funds, regardless of whether the company is actually doing anything better.

[dirck-norman]

Inflation adjusted 10 year cumulative return is ~192, no matter how much you don't understand how math works.

You don't re-apply inflation onto the principal.

[jacobgkau]

> Consumption has risen, inflation adjusted wages have risen for blue collar and white collar alike.

My wages haven't risen for nearly 5 years, while inflation has occurred over the past 5 years. Why the blanket statements?

> The main thing holding people back is the housing crisis. This is orthogonal to the value creation of businesses.

Are you suggesting a "housing crisis," in your words, wouldn't impact consumption? I'm watching my spending (and living like a child in his parent's house, except it's not my parent and I have to pay for it) in the hopes that in about a decade, I'll have saved up enough of a down payment for a home somewhere in my state that I could actually afford the mortgage on the remaining amount. There are plenty of things I'd potentially spend money on but won't as long as I feel like I'm economically stuck and have a chance in hell of saving my way out of it. So this feeling translates to fact.

If you think my personal experience is just an anecdote and doesn't count because it's not being told through the lens of large-scale numbers, fine. But I really agree with the person you replied to that you're gonna have to be a whole lot more specific than "value creation" if you want people to spend money on your AI products "in this economy," whether it's because they're actually strapped for cash or just pretending like you seem to think they are.

[decidu0us9034]

The upper decile of income earners account for more than half of all consumption in the US. Household balance sheets and wealth have never looked stronger, again when you account for all the appreciated stocks and properties owned by the upper quartile. True incomes for the lowest decile rose significantly for the first time since 1970 in 2022 and then sort of stayed flat again. Sure, statistically significantly, not "significantly" as in personally meaningful after figuring in rising consumer costs. There is a narrative where you can see all this as hugely positive but this is also largely a "vibes" based narrative. I don't know why you'd expect most people to care about what the "vibes" are like for the best off in society, that's a bit removed from their daily concerns.

[geraneum]

> Feelings aren’t fact... much of it has to do with rapid inflation and "continued vibes".

Oh the lost irony.

[dirck-norman]

Is it ironic? Or did you just read the comment incorrectly?

[WarmWash]

Sounds like internet sentiment and not research data.

It's kind of become socially taboo to not be suffering "in this economy", but on paper it's hard to see weakness in places that there isn't always weakness. As long as the 65-95% are doing well, there isn't going to be a collapse.

[forlorn_mammoth]

The most recent U Michigan 'Survey of Consumer Sentiment', which is THE authorative source in the US, shows consumer sentiment at the lowest levels since the survey started in 1977

From the U Michigan page: https://www.sca.isr.umich.edu/

or from the FED https://fred.stlouisfed.org/series/UMCSENT

[WarmWash]

Sentiment is just vibes though.

Everyone hates Walmart but they still do gangbusters numbers.

So people may be hating, but they sure as hell are also still spending

https://fred.stlouisfed.org/series/PCEC96

[jgbuddy]

A literal example is that I can use AI to file my taxes instead of spending a weekend and hundreds of dollars to have an accountant do it for me. It costs me like $5. that 245$ delta is the value of that output to me, as long as I am confident it is correct.

[mfuzzey]

Seems to be a thing in the US to need specialised software, an accountant or AI to file taxes.

In most of Europe individuals at least don't need any of that. I'm in France and it's just a connection to a government run website to enter a few figures, takes less than an hour most of it is already pre-entered (salary etc), the main thing to add manually is charitable donations.

If you're running a business then yes an accountant can be good (or be required depending on the legal form of the business) but not for individuals.

[bluGill]

In the US you don't need software or an accountant either. People have been convinced it is hard, but the reality is taxes are basic math for most people. It isn't hard to copy the few figures from a couple forms and add them up, the look things up in a chart. It is tedious, but not hard. That said, I use the software anyway because once in a while I make a human mistake and those are annoying when the IRS catches latter.

There are a few people who have a complex situation who need an accountant. Most do not, but since they have never looked at what is really involved they don't realize what is going on.

[asdff]

Taxes are one of those things that seem difficult and people reach for tooling or expertise without trying initially without, but are pretty easy to do yourself just filling out the forms.

[dgellow]

It really depends your situation. If you’re entrepreneurial, self employed, others things become messy fast, and you cannot just know what you need from filling forms

[asdff]

Luckily IRS makes this information available for you:

https://www.irs.gov/businesses/small-businesses-self-employe...

[moduspol]

Part of the value of paying an accountant is that you can get representation in case you are audited. Though I guess you did say you were confident it is correct.

[decidu0us9034]

This also sounds like rich people problems. Vast majority of people with a W-2 take the standard deduction.

[panta]

I think that to sum things up, we will have to wait until we can evaluate the cost of the mistakes. You could be lucky but you could also end up with a very negative output value in the longer time frame.

[WarmWash]

I did my taxes this year too with 5.5 and 3.1

Otherwise normally costs around $800 to do, because I have a small business too.

[smnc]

> as long as I am confident it is correct

Are you? Does it cost you extra (time or money) to be?

[jgbuddy]

Yes, and they were accepted. A year or two ago I would have been less confident but now almost UX is happy to cite sources.

[redfern314]

Not speaking to the wisdom of filing taxes using LLMs, but just FYI (assuming US here) taxes being accepted doesn't mean they were correct. It just means the IRS hasn't found anything major wrong (e.g. SSN used on multiple returns). Even being approved isn't a guarantee, an audit could come later.

[topaz0]

Even if an audit never comes they could be incorrect.

[deaton]

Thats the thing; the "increase in productivity" isn't being felt by the general public, the end user. If your "increase in productivity" just means more money being shifted around at the corporate level then it is meaningless.

[mrandish]

> There will be new value created by these models which people are happy to pay for which simply did not exist at all before.

True, but I think the GP's point was that what consumers will pay won't be nearly as profitable as what enterprises will pay to increase the output of their developers and knowledge workers. ChatGPT is currently the overwhelming leader in consumer AI usage but only ~5% pay $20/mo.

As a recently retired serial tech founder, I'm now one of those consumers. I use AI webchat daily for general search, Q&A and even to write little automation scripts for myself, yet I haven't paid anyone anything for AI yet. Even after being heavily restricted and performance nerfed to hell in recent months, free webchat AI is still fine for everything I do, and I'm not remotely price sensitive.

Even as AI compute costs fall over time, I doubt serving ads against AI webchat to consumers will generate the kind of high-margin, sustainable growth VCs get excited about. It's so undifferentiated I bounce around between all four leading providers because there's virtually no moat locking casual consumers to any chatbot beyond a single question thread. I guess if it had a nearly infinite context window seamlessly integrated across all sessions, that might be somewhat sticky for some consumers but it could also get creepy for some others - and it would devour gobs of the scarcest resource in AI. Beyond Maslow's Hierarchy of Needs, the mobile phone is the largest revenue, long-term mass consumer product ever but I just got a new flagship phone from a top-tier provider for $30/mo over 3 yrs. IMHO, even an all-you-can-eat, infinite context window, next-gen Mythos couldn't reach and sustain mobile phone levels of global consumer adoption at ~$20/mo. Unlike professional developers and knowledge workers, consumers don't have any "job to be done" big enough for an LLM to command that much of their zero-sum discretionary spend.

[fragmede]

What are the non-tech people in your life using AI for? $20/month, next to Starbucks and avocado toast, is discretionary. Maybe the novelty will wear off and non-tech consumers will leave it in droves, but everyone declared they'd leave YouTube if they started playing ads, but YouTube doesn't seem to have noticed.

[mrandish]

> What are the non-tech people in your life using AI for?

Mostly asking random questions they used to search Google for.

> next to Starbucks and avocado toast, is discretionary.

Sure, but your description implies highly affluent, urban professionals in western nations. I was talking about getting several billion global mass-market consumers to all keep paying ~$20/mo. Mass consumer adoption of mobile phones worldwide is currently >5.8 billion or >70% of humans alive. Only ~50M people are paying $20/mo for an LLM and I suspect many of them are not pure consumers but actually knowledge workers that AI vendors are losing money on and will eventually force into higher tier plans just like the $200/mo developers they're currently losing money on. These heavily subsidized loss-leader offers are all going away post-IPO.

Personally, I know maybe a dozen people who pay $20/mo for an LLM but only two of them are really 'pure consumers' who don't use it for knowledge work. Both of them are multi-millionaires and neither has had a job in ten years. One is retired like me and the other is so wealthy she has a Netjets credit card and has new cars delivered like some people order shoes. Everyone else I know paying $20/mo is a professional who uses the LLM for a lot of office or knowledge work and writes it off as a business expense - examples include a couple of attorneys who are senior partners in a law office they own, a solo architect, and a dentist who owns his own practice.

At $20/mo, AI vendors are probably losing money on most of my professional friends because they use it pretty heavily all day. They're only making money on the two multi-millionaires who both use it so infrequently they could easily be using free chatbots instead but are so rich they could lose $10,000 in their couch cushions and not notice. While they are profitable at $20/mo, they aren't exactly "typical consumers" that there are billion more of. I expect AI vendors will find ways to force my lawyer, architect and dentist friends to switch to higher priced plans soon because they're really knowledge workers abusing a consumer tier plan into unprofitability.

[jgbuddy]

100%, a driving factor will likely be how good we can make models that are so small they use almost no compute. Until then it is a race for adoption and moat-building (or screwing people over?) once you have users

[mrandish]

> a driving factor will likely be how good we can make models that are so small they use almost no compute.

That will certainly help but it doesn't move the fundamental limit because resource efficiency is a cost driver not a demand driver - and my argument is against the thesis that lying beyond professional devs and knowledge workers, there's an untapped trillion dollar industry serving LLMs to mass global consumers.

Using Simon's cost estimates, I agree that halving the current $1,000 - $1,200/mo MSRP to profitably serve frontier inference to professional developers and knowledge workers (PD&K) will help Vendor A steal share from Vendor B or C. It will also increase LLM sales penetration into the segments of the global PD&K TAM which can't afford ~$1K/mo for every seat. A fair chunk of the PD&K workers in many SMEs aren't included in today's ~$1K/mo per seat license pool, especially in 2nd and 3rd world geos. When the price falls to $500 and $250 most will but that's still just saturating the existing PD&K TAM - not pushing into mass consumers.

While the PD&K TAM is big, justifying Trillion+ dollar capex spend requires believing the TAM is much more than PD&K and eventually grows into converting a couple billion non-PD&K consumers into ~$20/mo subscribers. I don't buy it for two reasons:

1) The Comps: There are vanishingly few examples of long-term, mass consumer adoption of a discretionary technology at that scale. Mobile phones at ~$15 to $30/mo are the obvious one but LLMs are nowhere near being that valuable to the average plumber in Des Moines, baker in Jakarta or retired nurse in Hamburg. Pondering it, I just imagined forcing any of those people to choose between their mobile phone and an LLM chatbot. Sure, some who are flush with cash might choose both but for most consumers in the world ~$20/mo is big enough they'd have to pick one and ~zero percent would choose the LLM over their phone. After mobile phones, the second comp for discretionary tech spend I thought about was XBox and Playstation monthly gaming subscriptions but combined they have less than 90M paying subscribers and the ARR is just under $10/mo. As an industry, "Big LLM" is spending well over a trillion dollars every five years. XBox and PS ARR doesn't even cover paying the interest on that capital, much less the 3 to 5x returns hedge fund investors are betting on.

2) The Alternative: It's useful to doubt my own intuitions and one counter to my skepticism is to assume "But LLMs aren't finished yet, they're going to get much better." How much better could an LLM which can be profitable at ~$20/mo get than Claude Mythos in the next five years? Instead of debating future unknowables with myself, I've found it's better to just imagine the most perfect future product I can that's still realistically plausible. So, let's imagine we're willing to spend a million dollars a month to very unprofitably deploy a prototype to test the consumer demand for "Tomorrow's Awesomest $20/mo LLM" today. So we gather a few hundred super smart, broadly knowledgeable intellectuals together at one top-tier university research library, where they'll have access to every commercial database and unlimited Claude Mythos 2.0 and ChatGPT 6.0. Since our experimental budget is $1M/mo we can afford to add in several Nobel prize and Fields Medal winners too. They'll work together manually reviewing and improving not only every LLM answer but also our test user's prompts - and of course our test chatbot will have human-level real-time speech recognition and vision (via Zoom and screen-sharing with actual genius-level humans), making this truly a test of the "smartest, most accurate, best consumer chatbot" we can imagine.

Now, let's run the test by having one thousand mass consumers try it out and see how many Des Moines plumbers, Jakarta bakers and Hamburg retired nurses we can convert to a 1 year @ $20/mo subscription for our $1M/mo ultimate chatbot simulation. Playing this thought experiment out in a bunch of ways, I find some percentage of outliers, iconoclasts and closet intellectuals would go for it but... the vast majority just don't find it enough better than "free" chatbot alternatives AT&T includes with their phone subscription or Samsung bundles with Galaxy phones - despite only being ChatGPT 5.4-level. It turns out, most plumbers, bakers and ex-nurses don't have a compelling "job to be done" in their daily lives that even an MoE panel of actual Nobel and Field's medalists with ivy league professors can make enough more valuable than an inferior but free-to-me chatbot, in the judgement of our Des Moines plumber. While the world's smartest chatbot is nice, when it comes time to pay, he prefers having one additional premium football match on TV and a six pack of cold beers every month.

[fn-mote]

I'm having a hard time understanding this huge post that doesn't talk about enterprise users. I'm convinced that the consumer isn't going to be coming up with enough money to justify AI valuations... but doesn't this just mean that we expect the money to come from large enterprise users?

A recent post here said AI spend could be "20% of every software developer's salary"... and that seemed plausible based on productivity improvements. That's not about a phone bill.

[mrandish]

> doesn't this just mean that we expect the money to come from large enterprise users?

Yes, if you agree that not enough consumers will value "smart paid chatbots" over dumb free chatbots at ~$20/mo, then, as you say, the money has to come from enterprise developers and knowledge workers (PD&K). The big problem with that is the numbers don't work. There aren't enough PD&K workers that are highly-paid enough to justify $12,000 to $16,000 a year to cover the astronomical spending run rates.

Without recapitulating all the various scenarios AI CEOs like to hand-wave, my take-away is the scenarios which show today's frontier AI vendors earning the returns they've promised investors to get those trillions of dollars they're spending ALL require truly extraordinary deltas far above current historical actuals. Whereas ALL the scenarios which I find more plausibly realistic, even if still quite bullish, never even get near the required ballpark.

I've tried to make various scenarios fit but inevitably, when the PD&K demand side starts getting implausibly inflated, I start pushing the cost side down to keep the adoption rates remotely plausible. For example, assuming things like the number of PD&K workers will grow at 2X the highest rates ever seen before or that the percentage of the tasks knowledge workers actually do at sufficient frequency to really matter and which are also "LLM improvable" is implausibly vast.

It gets challenging because one quickly hits finite limits on the increase in tangible economic value that an LLM can possibly deliver on many common knowledge work tasks like drafting an email or a product proposal. Even if we assume next year's frontier LLM is so good literally ZERO slop is even possible so that our knowledge worker doesn't ever have to review anything - turning a 20 minute task into 2 minutes, or even 200 milliseconds, yields finite economic value to the enterprise. Even if you go extreme and suppose pretty crazy stuff like 18 months from now LLMs have eliminated 50% of PD&K workers, that messes with the spreadsheet assumptions because now the there are half as many $16,000/yr seats so the LLM license prices have to double again just to stay even.

At the edges it quickly gets nuts. I even tried assuming that in five years LLMs eliminate 100% of non-managerial PD&K jobs. All actual work is done by Super-AGI LLMs. Even if we assume such amazing intelligence can be profitably sold in five years for only 4x the price of today's far dumber LLMs, that Super AGI still costs $50,000/yr per human replaced. Even that assumption only cuts the blended labor cost of non-managerial roles roughly in half (depending on industry and geo). In the end, no matter how much labor cost LLMs enable companies to cut, it still only reduces overhead. Lower costs might allow Vendor A to steal some of Vendor B's customers but it doesn't increase the total TAM or demand of the entire sector both vendors serve.

Once you're out of plausible labor savings, one has to move to assuming that within five years Super-AGI LLMs working unsupervised will be making and validating fundamental scientific breakthroughs, then reducing those breakthroughs to engineering practice enabling new technologies which create entire new markets. Then they'll create whole new companies with maybe only 1/10th the humans and profitably grow those new markets. It quickly starts to feel like bad sci-fi instead of plausible, near-term financial scenario planning. Is stuff that extreme and unprecedented actually possible? Sure, incredibly unlikely and unprecedented things DO sometimes happen but things so far out of the distribution are very rare. But when making Sam and Dario's scenarios actually 'math out' starts requiring more than one "Black Swan" unlikely, historically unprecedented, earth-shaking event. And then goes on to need whole flocks of Black Swans, it just isn't credible. That said, I do believe that some parts of these scenarios may happen in 10 or 20+ years. It's just that the investment theses trillions worth of our 401K's are sunk into assume five or six year amortization and financial payback.

[PunchyHamster]

But will they pay big actors running top end models for that? You don't need latest openai or anthropic model to go thru your mails, get summary of the some products from web, or to do your to-do list.

The AI might very well be used by noticeable % of population daily, but that doesn't mean they will be paying trillion dollars to the leading US AI companies

[Planktonne]

> There will be new value created by these models which people are happy to pay for which simply did not exist at all before

What sort of new value, and why will people pay for it from someone else rather than prompting for it themselves?