Why OpenAI Is A Terrible Business
46:56 Share

T M O experience wherever you are.

model offers a huge selection of the style games with free spins, exciting promotions and always a you can waste your time with the others, or you can win at model register today at .

M O D O dog U S, for your free welcome. Bone moto was a little necessary. No, the following ad is sponsored by .

pets best insurance services. Your pet is your best. T, your thr, your preferred match. It's easy to love them even when they sneak your snacks. It's easy to protect them to with pentuer coverage from pets best, because it's all fun in games until they show on something they shouldn't with perfect timing. Best, best helps protect your very friend and your budget from this imperfect world. Up to ninety person on eligible that bills for less than a dollar a perfect match a pets best dot com peinture ance products offered a minter by pet insurance services are will see a under written by american peinture ance company or independent american insurance company for all terms, visit the best to come back policy.

One in three women and one in four men experienced domestic abuse in their lifetime, and nearly half of survivors delay leaving because they can't bring their pets with them. Parents started the purple leach project to help eliminate one of the many barriers domestic abuse driver space, a lack of pet friendly domestic violence shelters through the purple alease project perennis, helping to create more pet friendly domestic violence shelters across the country, so abuse survivors and their pets can escape and heal together. Visit perino comes lash purblow .

to get involved. The all electric chevy equal xv combines everything you need when you're ready to go. EV2EQ gives you an 是 of all electric range safety features for piece of mind and effort list technology， including the seventeen point seven eight diagonal display screen, all at a Price you love.

The chevy equal xiv is the first to drive all electrical C, S, U, V. That gives you what you need to do exactly. The manufacture suggested we tell tax license deficient open tiler set final Price to learn more, visit heavy dog comes flash equal X, V.

The best vacation stories happen when everything feels effortless. As america's favorite vacation company for fifty five years, booking with apple vacations is like turning on easy mode. You can find the best deals all on one side and enjoy added per.

So all that love to do is let the stories unfold, turn on the easy mode, and start your story today with apple vacations. Visit apple vacations. Stop, come or call your local travel advisor.

Cause of media one, welcome to Better off line.

As ever, please check the episode notes for links to the things that i'm talking about. Now frequent listeners know that I immense the skeptical of general AI, both as a technology but also was a business. It's expensive.

It's unreliable. He doesn't actually do that much. And the real world use cases ranged from kind of helpful to unhelpful, to not really useful in any way, shape or form, and definitely not unuseful enough to justify hundreds of billions of dollars in spending.

And I know somebody who's gna email, say, but had my mates, granny, uncles, dog uses check G, B, T to brainstorm or something else equally flimsy. And I really need you to put all of this in context. OpenAI loses five billion dollars a year, and basically every big tech companies blown past their emissions targets with few signs that they're ever bothered to try and meet them again, all in pursuit of the pit list. Almost convoluted use cases in history is a god damn fast. And that's why generate AI is such a regular topic on this show.

I feel like the tech industries experiencing a kind of collective matter, a delusion that seeing companies like microsoft and google, but both the companies and all planets, on a technology that continually fails to deliver on, well, anything I wanted say here, and I actually wrote down this supposedly massive potential, but what was the potential ever? Can anyone actually tell me what this was meant to do other than, anyway, eventually I see this all falling apart, and I think it's going to be a calamitous event when everything does actually unravel. I return so often to generate A I, because I truly fear the damage that this confrontation will cause.

I am concerned about the damage that's happening right now. We're from honest and want my tones off in a server. And I don't think you'd have A N A.

You are the way I don't really take much joy in watching these things burn. Sure, right. Mid is pretty funny thinking about such an a delegating fired or send up a shy and fired, both of them looking stupid. But it's not the same thing. My gleese kind of tempered when I stop and think about what all of this means or will mean.

I should say for the companies trying to get funding right now, for the companies that will try and raise money after the bubble, and for the take workers that will get laid off, either because microsoft, e and google want to put more money into data sentence in GPU or because the generative A I bets have gone, hits up and they need to save money some other way. And so I hope you will forgive me for someone more some but tone to this episode. I want to, again, make a clear headed analysis of where we are today and where I think we're going.

And the reason i'm retreats in this path is because over the last month or so, we've seen some really, really alarming signs about accelerate in crisis in generate AI and whispers that the party will soon come to a halt. Now you may remember a few epo des back. I started talking about pale horses at the AI apocalypse, signs that things were falling apart, like layoff S A I companies, Prices increases or decreases, internal discord, these big eye companies, speciously impressive yet kind of flimsy product announcements and so on is so forth.

Since then, i've kind of been proven right several times. Multiple pail horses have emerged. There was a big, stupid magic trick to impress investors in the form of open a eyes rush launch of its own one.

And that model, by the way, was code named strawberry rumor. Price increases for ChatGPT in the future. Now they're looking at forty forty dollars amount by twenty thirty, and they are looking at potentially thousands for models in the future and then layoffs.

Scale eye, which is one of the biggest players in AI training data, if not the biggest. And these are all signs that things are beginning to fall apart. And I think it's important to explain how precarious things are right now despite all of the money stock ing around, and how dangerous the power of magical thinking really is.

I want to to express my concerns about the fell of this movement and the obsessiveness and direction list that brought us here. And I want us all to kind of do Better. Now I don't mean you to listen up.

If you remember of the media is written something vague about A I and you have kind of Carried water for sam altman. IT is time to stop. IT was already time to stop some time ago, I should just be clear, but now really is the time to stop.

But this week are two party, and i'll be explained things well in two parts. Obviously, first of all, we're onna talk about open eye, the most recent funding round and how bad their businesses and exactly how worried you should be about them. Now the second part is going to be about something i'm calling the sub prime a eye crisis where I see a growing bubble of companies integrating generate AI at Prices, well, heavily subsidize by big tech and heavily discounted by companies like OpenAIron an thropic.

Should either of these companies get desperate or, I don't know, actually want to make more money than they spend, they're inevitably gonna have to raise the Prices to match their actual costs, which will likely make the existence of many A I startups completely untainted. You know, like when the teeth of interest rates on sub prime mortgage that expire during the financial crisis and people suddenly started losing their houses. In any case, these are gonna be meaningful year kind of brutal episodes in part because while it's enjoyable to watch big tip burn or get embarrassed, there's a real human cost, like I said.

And the damage to our environment is already being done and might not even stop when generate V I does. And whether microsoft and google in the other big generate AI back has slowly wind down their positions or cannabis ze their companies to keep OpenAI on and anthropic alive. And convinced that that the end result is still going to be the same, I fear tens of thousands of people lose their jobs and much of the tech industry will suffer.

They realize that the only thing that can grow forever is a cancer. I'm going to paint you a bleak picture, not just for the big A I players, but protect more widely and for the people who work at these companies, and tell you why I think the conclusion to this sorted saga, as britain damaging as IT will be, is coming sooner than you think. Let's begin.

As of explained in agonizing detail in the past, in my newsletter, OpenAI, e will have to continue to raise more money than any startup has ever raised in history in perpetuity, just to survive. They're gonna burn five billion dollars this year. And yes, that's after they make their revenue and costs are only set to increase over the next few years, they develop bigger models.

Their business as IT stands, is completely untangle. And i'm going to explain why I A bit nevertheless why OpenAI the extensible non profit that may soon become IT for profit, ladie knows at this point raised the funding round evaluation of one hundred and fifty seven billion dollars, bringing in the report at six point six billion dollars in cash, along with opening a four billion dollar revolving credit line from banks like J. P.

Morgan city and a bunch of others that should know Better off the acquisition of twitter. Investors in the funding round include josh cushion's, thrive capital in video and of course, microsoft, who was somehow undeterred by the abbot tax itis, a cco mirror ity days before the round closed, also joined him, the free. He was soft bank, the japp ane's investment company, with a drug record of making some of the worst bets in history.

And they dumped five hundred million dollars into open a eye. And that alone should stop getting people little worried. Yes, a soft banks found a massage.

Shi son is what we in the business call the dumas. I realized this is a very broad brush to paint somebody with. But let me explain. South bank vision fund, sixteen billion dollars into we work, which by the end of last year, by the way, was worth forty four million dollars and nearly a billion dollars into ware card, which turned out to be an elaborate fraud.

With its leadership of the face in criminal charges of becoming future tips from justice in russia, soft bank has become effectively anonymous with burning cash on terrible beds and lost over thirty two billion dollars in the last three years. The problem the opening eye faces is that they need an absolute shit on our money. And generally, the only companies that need that much are the ones that may not have a sustainable business model.

OpenAI demanded a minimum investment of two hundred and fifty million dollars from investors. And while regular american vies might have that much and indeed had, in the case of costliness coati management, there aren't enough of them to fill out to six point six billion dollar funding round. And because open a ee needed that much, not wanted, needed.

There are only real choices were to go to maco shi son, whose a man who put tons of money into stupid things in believe he was put on this earth to make digital super intelligence and have a linked to that, by the way, it's insane. Well, I mean there is also another choice ah you know the vados your investors in the middle east m ter M G X, A hundred billion dollar investment fund backed by the united arab ts to invest primarily in A I and semiconductor companies. This should also be a big warning sign that things are going poorly because absolutely nobody raises from the U A, E.

Of the south because they want to there the place you go, if you need a lot of money and you're not confident anybody else is gonna IT to you. Other aspects of the deal are also a bit worrying. One of them being that one of the founding partners of M G X is the sovereign wealth fund of haidari, which brought in around five hundred million dollars into anthropic in twenty twenty three.

Kind of a conflict of interest. You think any of these people actually got them care, though they're all working to the same raw economy nonsense? Anyway, as I mentioned previously, OpenAI now has access to that four billion dollars in debt from banks in that revolving credit facility to and that's also not brilliant for a company that just burns cash.

I don't know what the interest rate on IT is, but I can't imagine it's great. Generally, revolving credit lines are worse interest rates. So that's great, I guess.

But open the eyes, desperate fund raising comes down to one one thing, one really obvious thing. They need so much money. They needed black ten billion dollars y and they kind of got IT in this one key way.

But they need that money to survive, because, like a set, they lost five billion twenty twenty four. And that figure is likely to increase as more complex models demand more compute and more training data. And anthropic C E O dio a day predicted that future models may cost as much as a hundred billion dollars to train and open a eye.

By the way, they may have just raised the biggest round in history. They're gonna have to raise another round again soon. And i'll explain why. While OpenAI succeeded here, IT was only after fairly August process where the'd already tried to raise a one hundred billion dollar valuation earlier in the year, and that specifically turned off investors because of the huge Price tag. And there's this, by the way, to quote the information, there's a growing concern over the overvaluation of generate AI companies.

How would you fuck and think do you think that there's a concern here? Well, after this podcast, you could have be really, really concerned, I guarantee, anyway, to get around done. Open eye, committed to converting itself from a nonprofit to a four profit entity, failure to do so within two years, i'll see their funding converted into debt, which, by the way, will be the kiss of death for a company that only ever loses money.

The deal per voters also includes provisions that would allow investors to adjust evaluation, claw back their investment should the transition from non profit to four profit fail. And I mean, this is worrying because I don't think there's any historical presidents of a company a on profit at least that had this much money than converted. And more complicate that as well, is converting a business of this scale from a non profit to a four profit well involves transfering assets.

And because any assets previously donated to the public benefit, as Alexander read partner at baka hole, Stella is quoted as saying in the wall street journal, could not be converted into a private benefit without compensating the public for the loss. Every asset, which includes things like patterns, another intellect al property, would have to be paid for. While it's hard to save for certain, it's likely this deal in the process of untangling open eye from its nonprofit beginnings.

He's really only going to add to open an eyes cash problem, and it's not like that was a good problem to deal with in the beginning. And now I will do a thorough and anal sis of OpenAI, because I want you to see exactly how stupid things that become in the tech industry, how ridiculous all of this writing. This boiled my blood, by the way.

Because when you see how lossy this company is, when you really ingest how bad open a eye is, and how stupid IT is, that we've let this happen and we, meaning the tech industry, you then will think on your own life and be like wall. You think I could pay my rain, like lay, or like ask a company to pay my rap? Of course, my job involves me just burning powers of cash.

No, you wouldn't be able to lose money for a business. You and I, we, we have to Operate Normally. Not sam mult man, though. Sam mult man doesn't have to be Normal. Sam molton like all of the guilded ask holes of the tech industry, is allowed to live in this fantasy land based on marketing high. Now before we go into IT, though, i'd like to enjoy one of the following lovingly crafted advertisements in no way see out of place from anything i'm saying.

This is an a tiny section with you hear, because of our job, we need to be connected, tony, over connected by more than us.

But it's important like .

connection with your friends, your family. Every connection enriches your life, and that gives you something that provides. And high speed internet, just in the context of this show, is so important, we will be talking about something here.

And it's easy to forget when you have IT. You know, big parts of the country still don't have high speed internet.

So A T N T knows this. They're ware of IT at A T T, and they're making a huge effort to change that. What you're doing is on trying to come thirty million plus locations currently who don't have IT to get IT with fiber by the end of twenty, twenty five.

Millions of people obviously will benefit families, businesses, schools, hospitals. There is a place called old county in kentucky, where they're now providing high speed internet to more than twenty thousand customers. A T N T join a lot of great things.

Connecting changes everything. A T N T.

even if you think it's a bit overhyped, A I is suddenly everywhere, from sell driving cars to molecular medicine to business, if it's not in your industry yet. IT is coming fast, but AI needs a lot of speed and computing power. So how do you compete without cost spiralling out of control time to upgrade to the next generation of the cloud? Oracle cloud infrastructure, or O C I.

O C I is a blazing fast and secure platform for your infrastructure, database and application development, plus all your A, I and machine learning workloads. 好起来 cost fifty percent less for compute and eighty percent less for networking。 So you're saving up pile of money.

Thousands of businesses have already upgraded to O, C, I, including M, G, M resorts, specialized bikes and firework ai. Right now, oracle is offering to cut your current cloud bill in half if you move to O, C, I for new U S, customers with minimum financial commitment offence. December thirty first, twenty twenty four. See if your company qualifies for this special offer at oracle dot com slash strategic. That's oracle dot com slash strategic.

Did you know your life insurance is an asset you can sell? M. J. Jackson, CEO of advocates. For almost twenty years, advocates has been purchasing life insurance policies from seniors just like you, and IT advocates pays that com.

We've created a free policy value calculator so you can find out the worth of your policy in just seconds. There are no fees and no obligations. Get real value for your life insurance when you need IT with advocate, learn more. Advocates.

pays doc. Experience wherever you are.

model offers a huge selection of the style games with free spins, exciting promotions and always change jackpots. You can waste your time with the others, or you can win at model register today .

at M O D O dog U S. For your free welcome on mois, a cesspool. The following ad is sponsored .

by pets best insurance services. Your pet is your best, your thist, your preferred match. It's easy to love them even when they sneak your snacks. It's easy to protect them to with pentuer coverage from pets best, because it's all fun in games until they show on something they shouldn't with perfect timing. Best, best helps protect your very friend and your budget from this imperfect world up to t 5 on a live bills for less than a find your perfect match， a pets best stock com p insurance products offered a mister by pets best insurance services, I will see are underwritten by american peinture ance company or independence american insurance company. For all terms, visit the best to com back policy.

And would back. So to put IT bluntly, open a eyes and absolute dog of a company. On september twenty seventh, the new york times reported the open eye would lose five billion dollars in two thousand twenty four, a number which the information is estimated back in july, and the OpenAI expected to raise the Price of ChatGPT, plus its premium product, to twenty two thousand a month by the end of twenty twenty four and remarkable forty four dollars a month by twenty thirty.

Interestingly and worrying ly, the times also confirmed another hypotheses of mine that, and I quote, fund racing material, also signal the OpenAI would need to continue to raise money over the next year because its expenses grew in tando with the number of people using its products. In simply terms, OpenAI may have raised six point five or six point six billion funding literally a week ago, but they need to raise more, probably the same amount or more, i'm going to say, within the next six months. And the times also reports the open eyes making internal revenue estimates I would describe as this is a technical term, absolutely fucked in ridiculous open eyes monthly revenue hit three hundred million dollars in August, and the company expects to make three point seven billion dollars this year.

The company will, as mentioned, loose five billion doors. Anyway, I am not going to keep reminding you of that yet. Open our eyes says that he expects to make eleven point six billion dollars in twenty twenty five and stonishing hundred billion dollars by twenty twenty nine, a statement that is so egregious that i'm surprised it's not some kind of financial crime to say IT for some context by the way, microsoft makes about two hundred and fifty billion dollars.

Google about three hundred billion dollars a year, and apple about four hundred billion dollars a year. So yeah, I guess by twenty thirty, that's how big OpenAI talk drives me insane. And also, by the way, all of those companies make more money than they spend.

Open an eye, on the other hand, spends two dollars and thirty five cents to make one dollar. If you remember, one thing from this podcast is that every dollar they make, they have to spend two dollars and thirty five sense to get IT. IT drives me insane.

People, we don't have to live in these conditions. How do I do this? How do I get to burn money? The no.

Seriously, OpenAI loses money on every single transaction, every single time that somebody uses ChatGPT or plus connect one of their models. And while they might make money selling previous subscriptions, I severely doubt those subscriptions are turning a profit. And now I really do think losing money even on their power users, and I will get into the next episode.

I believe there's also a sub prime eye crisis bringing because OpenAI API services, which let people integrate their models into products, are currently Prices to attract customers and scale. And increasing these Prices to match the actual cost will likely make this product unsustainable for many businesses currently relying on these discount rates and that if they have any usage. But i'll also get to that later.

Now if you anything like me, you have somebody told you the OpenAI is a growth business and then IT will just turn the nob to make itself profitable, much like amazon web services, which is amazon's cloud computing product years ago. Now I just want to be clear, if you're listening and you think that you just wrong, you just completely wrong. You live in in a fancy land.

Wake the fuck up. I'm sick, tired of here in this crap. open. I is nothing like amazon web services. First of all, open eye owns none of its infrastructure as everything everything is wrong of microsoft as your cloud.

Secondly, microsoft also, as part of an OpenAI funding round from twenty eight, has full access to all of open, a paid research, which is all of IT, by the way, and full license to cell and integrate all of their technology. They have complete free rain over open their eyes. Intellectual property.

Thirdly, amazon web services did not immediately start with a deficit, so greatly required raising more money than ever been raised in the history of tech, where still open, a technology doesn't get cheaper as its scales. In fact, IT does the opposite. This is not like other businesses.

You want to compare this to stop saying this to people. It's ridiculous. And I know why people say this and I am ranting, and you just have to forgive me.

I know that people don't want to believe that this much money can be wrong. IT can be IT is currently being wrong. You are watching and listening to IT being wrong.

This is not like google cloud. IT is not like the early days of the internet. IT is not like amazon web services.

It's nothing like them at all. There is no historical example like this. Not even uber.

Uber at least had a business model in the worst year of its life. Uber lost, I think, six point six billion dollars or something like that. So about one point six billion dollars.

No en A I lost in twenty twenty four, right? You want to know what year that was, twenty, twenty, the year when people couldn't go outside. And there are many years after the troubles should have stay in those too.

But that's a different podcast. The point i'm making is if you've got a comfortable little thing in your soul that saying this is just like things i've seen before, they'll turn IT around, they absolutely warm. And now i'm gna break down exactly how untenable OpenAI is with a few statements.

Frustrated OpenAI e is trying to hit eleven point six billion dollars of revenue by the end of twenty twenty five. And to do that, I will have to trip its revenue and its current cost of revenues. So two thirty five, to make a dollar open A, I will spend twenty seven billion dollars to hit seven point six billion dollars, even if open a eye somehow harms their costs.

And by the way, in the next year, they are going to be building a brand new model, which your customer absolute shit ton of money, they'll still lose two billion dollars in twenty twenty five to hit these revenues. However, I must be clear, these costs are going up. They need this company to grow by three hundred percent.

And the only way they're going to do that is by making a new model. And the only thing that's really growing at these companies there three years is and those three years is lose the money every single time they use them. And even if they added a two dollar Price increase to ChatGPT plus and a similar Price on their business plans, that teams and enterprise, these things aren't really gonna significantly move the needle without also adding a bunch of growth on top and open your eyes.

Latest GPT four model IT costs one hundred million dollars to train and more complex models like the iran when they are allegedly coming up with. And I imagine o one which will get to next episode, are going to an absolute shit. More to build.

And also, by the way, the information estimated back in july the OpenAIce tr aining co sts wo uld be ab out th ree bi llion do llars in tw enty tw enty fo ur. The question are coming down. People, now here's s another problem, buzz.

Tech companies need exciting things to get investors in customers jazz and opening. I will ever really had anything exciting or truly important since the launch GPT three point five and even the latest reasoning model luck is it'll get to the next episode. It's not been particularly impressive.

That model, by the way, is much more expensive to run and uses something called chain of thought reasoning, which i'll get into. And ethel IT takes more power to run because they're doing way more calculations and they didn't even have a use case when they announced IT what is going on with this company. But put in that aside, OpenAIce pr oducts ar e al so be coming in creasingly co mmoditized wi th go ogle ma tter on an ev en mi crosoft bu ilding th eir ow n la rge la nguage mo dels an d ot her mo dels to co mpete we ll.

Still, these models are using effectively identical training data, which they're also running out of, which makes their outputs and by extension, the technology itself kind of similar, but most worrying of all. And i'm really gna drumm on this one later, open an eyes cloud business. So the one will they connect.

They are models to other businesses and then the business is sell products to customers using those models technology. It's small. It's really small. It's small to the point that IT suggests there's a fundamental weakness in the generate A I industry. It's extremely worrying that the biggest player in the game only makes a billion dollars less than thirty percent of its revenue from selling access to its supposedly innovative technology. And fundamentally, I can't really find a he compelling evidence that suggests the open aee is gona be able to sustain this growth.

In fact, I really can't find any historical comparison for this company, and I also feel like open a eyes, growth is already stumbling, but if you don't want to stumble through life perf of joy or growth or happiness, I really recommend you buy one of the following things. If you can't buy IT, download IT. If you can't download IT, go into a shop and demand that they start stocking IT. And if they're just give IT, you demand that you can give the money. Listen to the advertiser.

Did you know your life insurance is an asset you can sell? M. J. Jack and CEO of advocates life for almost twenty years, advocates has been purchasing life insurance policies from seniors just like you. And IT advice pays that com.

We've created a free policy value calculator so you can find out the worth of your policy in just seconds. There are no fees and no obligations. Get real value for your life insurance when you need IT with advocate, learn more.

Advocates pays 点 com。

M O D O die U S. For the best replace social caso experience wherever you are, model offers a huge selection of bigger style games with free and exciting process and always a jack. You can waste your time with others, or you can win to register today at M O O U S for .

your free component processor distance web. The following ad is sponsored by .

pets best insurance services. Your pet is your best day. Your thist, your preferred match is easy to love them. Even when they sneak your snacks.

It's easy to protect them to with pennicut verage from pets best, because it's all funning games until they chew on something they shouldn't with perfect timing. Best, best helps protect your very friend and your budget from this imperfect world. Get up to 20 on that bills for less than a dollar a day。 Find your perfect match. A pets best dot com petition products offered a minor by pets best insurance company.

The all electronic chevy equal X E V combines everything you need when you're ready to go E V, starting at thirty four nine ninety. The twenty twenty five equal X E V L T gives you an impressive baLance of all electric range safety features for peace and effort list technology, including the seventeen point seven eight diagonal display screen, all at a Price you love the shaver equity found to drive all electric S U V that gives you what you need to do exactly you the retail Price school tax license deficit of equitius er set final Price to learn more visit chevy dog comes flash equal .

ox v say goodbye to winter blues and hello to Sunny views this holiday season. Apple vacation is making IT easier than ever to escape to the destination you've been dreaming of, whether is the White sense of the carribean, the Crystal clear waters of mexico or the beauty of the we've got you covered book now and enjoy up to four person off at the hottest resort. But don't wait, these deals won't last long.

Head over to apple vacation stock com or call your local travel agent today. Pack your bags. Paradise is calling and it's waiting for you.

Alright, with back. So for the next part, this episode, i'm going to digg into exactly how OpenAI makes money, which is going na require me to lay out its various businesses and products and it's going to have a lot on numbers. Trust me, it's worth IT because this sheet is insane.

So according to the new york times, OpenAI expects ChatGPT to make about two point seven billion dollars in revenue in twenty twenty four, with an additional billion dollars in revenue coming from other businesses using open a eyes technology, roughly seventy three percent or two point seven billion dollars of OpenAIce re venue co mes fr om se lling pr emium in versions of ch at PP T ca lled pl us te ams an d en terprise Ch atGPT pl us ma rketed to in dividuals fo r tw enty bo x a mo nth, offering fast the response times priority access to new features and capability is not found in the free product like image generation, which you can get in so many different of the places. Importantly, open eye can use anything you do as training data unless you explicit up, down. OpenAI e also sells access to the teams product, which cost twenty five dollars a user a month of piani a three hundred box a year per user and thirty dollars a user a month of paid monthly.

From this point on, your data excluded from that used to train OpenAI models by the form OpenAI sells enterprise subscriptions that, including expanded context, went over longa prompts, meaning you can give more detailed instructions, admin controls and in hunt support an ongoing account management. IT isn't clear how much this cost, but I found a redit thread from a year ago that suggests it's about sixty dollars a user a month with the minimum of one hundred and fifty seats on an annual contract. I also believe open an eye off his discounts for customers to buy a bulk, very standard software as a service business model.

A further billion dollars, or twenty seven percent of open a revenue comes from OpenAI licensing its models and services via its A P. I. One thing you know is when you look at this Price page is that there's this huge variety of models nap is available and that there's massive amounts of variation in pricing too.

I look into these in depth in my newsletter, which bother is called open aee is a bad business and it's worth reading if you want to, to know if not just for the fact that the pricing in its various services also super messy. But you don't need to know that right now, open eye also makes about two hundred million and a year setting access to its models through its microsoft, according to bloomberg, where OpenAI takes a two twenty percent cut of all revenue. And by the way, that's not on top of the billion.

This means that OpenAI only makes about eight hundred million dollars a year by selling access to its A P. I. That two hundred million is on top. I want to add how worrying this is. Both OpenAI and the larger generate A I market.

If OpenAI, the most prominent name in all of generate A I, is only making a billion dozen a year from selling the shovels for the gold rush, what does that say about the growth trajectory for this company or the actual usage of generate vi products? That's a question for late in the show. The first I talk dollars.

So as IT stands, open eye makes the majority lucky, said more than seventy percent of its revenue from selling access to premium versions of ChatGPT. A few weeks ago, the information report that ChatGPT plus had more than ten million paying subscribers, and then I had one million more than what paying for higher Price plans for business teams. As I ve laid out previously, this means that open your eyes, making about two dollar million dollars at a month from consumers subscribers.

But business teams is a very vax IT isn't obvious what IT means. That could mean the team plan, that could mean the enterprise plan. And you may think, well, ed, couldn't IT be horribly wrong. Couldn't IT be that there actually ripping they'd got got a million enter Price subscribes wouldn't be good wrong though.

The new york times has been back here based on their reporting, we can actually get a little more specific ChatGPT, plus ten million customers making OpenAI around two point four billion dollars a year, ten million users spending about twenty box a month, which equites two hundred million mark plied by twelve, two point four billion dollars. Dad, I swear, I was listening a math sometimes. This means, by the way, that business uses make up about three hundred million dollars a year in revenue, or twenty five million dollars a months.

IT isn't really great. Well, ten million paying customers might seem like a lot. ChatGPT is effectively to generate A I what google is to search ten million people paying for this counter table stakes.

And the idea that this company can triple that number in a year is absolutely ridiculous. OpenAI has been covered by effectively every media outlet in the world. It's mentioned in almost every single conversation about A I, even when it's not about general of A I. And as the backing of marketing push of microsoft and pretty much the entirety of silicon valley, ChatGPT has over two hundred million weekly users and the new york times reports, the open eye has, and I quote, three hundred and fifty million people using their services each month as of june. That was unclear if that includes those using the API.

I would guess that ChatGPT users, collectively, this means the open eye, the most popular company in the industry and the most prevalent industry and the industry talked about to everyone and everywhere, can only convert about three percent of its three uses into paying customers. Now, by the way, eager listeners may go well. Three, twenty percent conversion.

That's not bad. right? right? wrong. Generally, free products don't cost this much to run. ChatGPT itself is extremely expensive. When if it's free or whether it's premium, it's the same thing. So yeah, there's three hundred and fifty million people that more of like a parasite than anything. The ninety seven percent and we won't give them the credit card is not great.

And I think that lack of conversion might be because ChatGPT and ChatGPT plus a really similar products, ChatGPT plus that you use the product more often and have access to nummer models. But there's no obvious new thing you can do as a result, you can't really up sell to plus unless you found someone who's already worked out the use case for themselves. An OpenAI remains peace poor at actually marketing this product, mostly because of the limitations of what a large language model can do.

As a result, most people did ling with ChatGPT will get what they need to out the free version. I'd also argue that those willing to pay for a plus subscription are more likely to use the platform way, way more than three years as and because every prompt on ChatGPT loses the company money, it's reasonable to believe that paying user would be far more of a burden on the system. While there's a chance the OpenAI could have a chunk of use is unparticipated tive.

One cannot run a business based on selling stuff you hope that people won't use. No, no, it's nothing like a jimmer insurance stop IT. When I wrote the news letter and I said that I got so many people thought they were the clever as people in all school always like a gym.

It's like, sure, it's not the same. Stop IT you not clever. No one's impressed. And I should be clear, there's also a reason why enterprise customers are generally more desirable than private individuals.

As with any other consumer centric subscription product, regular customers are far more likely to cut their spending when they feel like they're getting value from the product or when the household budget demands or their economic problems. Just look at netflix. They were the biggest name, while they still are the biggest name, streaming, and they lost a million customers in twenty twenty two because of the cost of living crisis.

These are real things that will happen to OpenAI if they're not already happening. Chat, P, T, plus is likely for many people, kind of a lifestyle product. And the problem is when people lose their jobs or inflation hikes happen, or cheaper things come along that do much the same thing, these products are the first to get slashed from the budget.

And as before, any arbitrary or silly or desperate Price increases done by a company, there isn't a good business. Honestly, it's kind of remarkable. The OpenAI found ten million people to actually pay for ChatGPT.

But how do you grow that to twenty million or four million people? I think they need like thirty million by the end, twenty twenty five. How that happened? Very boring anyway. At present, OpenAI makes about two hundred twenty five million dollars a month as two point seven billion a year. By selling these previous subscriptions to hit eleven point six billion dollars in twenty twenty five, open eye would have to increase revenue from ChatGPT customers by three hundred and ten percent if we consider.

And forgive me on when a dozen math at you, the current ratio of plus subscriptions, the teams and enterprise subscriptions, about eight eight point eight nine percent is GPT plus versus the business once eleven percent, OpenAI would need to find eighteen point two nine million paying users and that at the new Price point of twenty two bugs a months, while also retaining every single one of the current subscribers to ChatGPT plus, who would also need to renew at the same Price point to hit seven point four billion dollars, or about six hundred and sixty million dollars a month. IT would also have to make an additional nine hundred and thirty three million dollars in revenue from its business or rent Price client, which again would require OpenAI to more than triple their users and retain the current ones to triple these users. To actually do this, ChatGPT has the meaningful ly change, and I have to do so soon.

Or disclose multiple meaningful, powerful use cases are so impressive that eighteen million new people agree to given twenty two dollars a month. That's an incredible, and I would say, insane go. And one that I do not think this company is capable of.

Achievement IT would require making ChatGPT something far more compelling that already is in ways that I can't even imagine, just as the company lost multiple members of its senior leadership team, IT doesn't look good. There needs to be a meaningful change. Now I know i've already given you a lot of worrying things about open eye, about the very top heavy subscription driven business.

Why isn't gonna grow? Why they need to grow IT to grow fast and name one's ever grown, why they need to stop losing so much money. But there's something else to be worried about.

There's something even flimsier about this business, and it's actually something that's disastrous. It's disastrous. I'm serious. IT is simply disastrous.

How little of open a eyes revenue comes from providing other companies that means to integrate, generate vi into their systems. IT is astonishingly bad. I cannot be clear enough here before I wrote this script.

Before I wrote the newsletter that LED to him, I believed in my heart hearts that how OpenAI made their money was from providing access to their models. That makes sense, right? Because these companies talked about everywhere.

Their technologies is meant to be innovative. Everyone's meant to use IT, right? That makes sense.

Surely most of their money would come from people getting in on the revolution and then integrating the revolution into their products, then people using the revolutionary thing, right? wrong. First of all, if OpenAI only makes a billion dollars a year selling A P.

I access, and thus let you integrate the models into your products, IT suggest even the biggest company in generate A, I cannot find enough customers to make its business viable. Secondly, IT suggest there's a remarkable ably small amount of demand for generate AI integrations. Or consider another way that the companies connecting to OpenAI aren't really making you very much money at all.

If the generate vi revolution was really here, surely OpenAI, the household name in large language models, would be rolling in cash from their ibi business, not leaving IT at less than thirty percent of their annual revenue. This isn't like it's so strange. So at the twenty twenty four OpenAI dev day event is develop, a conference which took place in october.

First, the company said that over three million developers are building apps using OpenAInfrastructure th at wo rks ou t to ab out th ree hu ndred an d th irty th ree do llars of de veloper. And that's far less money than other companies which make money by providing a service through their O. P.

I actually make. Twila, a company that makes its money sending SMS messages and push onal fictions for companies over the past quarter, made about a billion dollars revenue. That's what OpenAI made from renting out its models.

A, P. S, over the past year. Toilet o also made roughly four billion dollars over the last four quarters, which is more than open a eyes project a revenue for the entity of twenty twenty four.

As I mentioned before, OpenAI makes two hundred million dollars of its one billion dollar revenue from microsoft reselling its models, a twenty percent cut that suggests that microsoft is making about a billion dollars from OpenAIce mo dels. So in the event that open a eye in microsoft making about a billion dollars and animalized revenue by a friend, access to open A S. Again, these are estimates based on current growth factories.

IT suggests that there's only two billion dollars in annual revenue coming from both these companies combined, and that's without way. Is OpenAI making less money from open in eyes models? And microsoft is, oh my god, this business is a thinker.

And this also suggests that generate vii as a technology doesn't have the product market fit, according to a survey by Andrew and horowitz, a VC, earlier in the year. And I quote, the twenty twenty three market share of close source models was estimate at a eighty to ninety percent, with the majority of that share going to OpenAI. And open source ones would include things like matters lama model, which is kind of open source.

That's a whole other thing. Anyway, another survey from IoT analytics published late last year suggested the number might look a little different with thirty nine percent of the market share going to OpenAI and thirty percent going to microsoft. Assuming that the latter numbers are true, even close the true.

This suggests that the the general of a high market is really small. If OpenAI, which dominates, we have a wager. About thirty percent of the market is only making a billion dollars a year from selling access to its models at this stage in this massive hype bubble that might not even be ten billion dollars of.

And your revenue from companies integrate, in general AV I into their products. That's tiny. And this should be where all the money is.

If this stuff is the future of everything, why is the revenue streams so painfully week open your eyes is making so little selling access to their models? And IT suggests that despite this hyped cycle, the rider is an interest in integrating these products from developers. Or when these integrations are actually in there, consumers aren't really into them or using them.

Remember, these products are charged on usage, and so it's possible for generate avi to be integrated into a service but not actually drive much revenue for open eye as a result of users not really caring all ChatGPT. As brand recognition companies integrating open our eyes models into their products are far more indication of the long term health of both the company and the industry. Self, because if open a eye can't convince people to be integrated, use this shit.

Do you think others are succeeding? I mean, think about that. Where if have you seen some weird chatbot appeared in your life? Where have you seen an L M.

Poked into something you use? Have you been like, all good, I can't wait. Or if you just kind of tried to ignore IT, maybe i've interfaced with IT.

IT sucks. There are some decent products like there are email clients that summarize emails, microsoft teams, apparently people really like the summer zone. But we are meant to be describing the future here.

We meant to be describing something exciting and sexy. Not, uh, this summer is the meeting for me interesting. But looking at these numbers, it's hard to imagine a open a eye will more than triple revenue in the next fifteen months at eleven point six billion dollars in sales.

Furthermore, at its current bury, OpenAI is currently spending, luker said, two doors and thirty five cents to make a buck, meaning that eleven point six billion dollars in revenue could cost as much as twenty seven billion dollars to actually make. And as a previously mentioned, while costs could ably come down, all signs point to the increasing GPT four costs one hundred a million dollars to train, and more complex future models will cost hundreds of millions of billions to train. As I mentioned, the information said the training costa would look like three billion dollars in twenty twenty four.

And I think it's fair to assume that the new model suggest as costly, if not more so. Yet, there's something deliciously ironic about all of this that despite a clear lack of user interest in generate A I open a house global marketing push, succeeded in making lots of people and tried enough to try a completely free service that only loses the money. But this usually great news that the product is three hundred or more million for users.

Every good time somebody uses a service like ChatGPT, IT loses the company money, and in the case of ChatGPT, good lord, they must be losing so much. The information estimated in july, the opening eye will spend around four billion dollars in server costs in twenty twenty four to run ChatGPT and host other companies run their services using GPT in its other models, effectively meaning that every dollar of revenue is immediately and by the cost of acquiring. And that's before you're factory.

More than fifty one hundred people, many is seventeen hundred people, now work up OpenAI, which is another one point point five billion dollars or more in cost and other costs, but away on top that they got real estate, ate taxes, stock grants, is this is all very bad. And while OpenAI could potentially reduce costs, theyve shown no proof that they can and they tried once. Well, at least one got out the so called iraq as model from last year that they tried to show that was meant to be more efficient if failed to launch.

That's not a good sign as IT. But there's one more problem. There's still more problem. And this one, well, this one was caused by their fund racing. You see they just raise six point five, six point six billion dollars in capital in one hundred and fifty seven billion dollar valuation. This means that all future rounds have to be at that valuation or higher.

A low evaluation, which called down around would make current investors quite psy and send a very loud signal to the market that the company is having trouble because nobody thinks they're worth what they used to be, which in turn would overwhelmingly suggest the open a eyes only way to survive is to raise its next valuation at a two hundred billion dollar valuation and require yet to another giant race, I would say, at least ten billion. For context, the biggest IP evaluation U. S.

Corporate history was alabama, which they build on the M, V S C, with a market cap of nearly hundred seven billion dollars. That figures more than double the run up up facebook, which had a value of eight one billion dollars. And I, anna, give you some history on that one.

Two, facebook's initial IPO was really bad because people are concerned about mobile users and not monetizing them. Well, they want that one, but the markets poverties facebook for that. Do you think that they're gonna be out yet so your company loses five or more billion dollars a year and you have no path of profitability? Yes, put you up on the next stack sounds perfect.

Fuck that. No, they don't going to do that. How is this company gonna do.

How is this company going to I, P. O? And yet there's more problems too. I don't know how open our eyes meant to convert itself from its weird non profit structure into a four profit company. I don't know if it's possible IT won't be easy.

And what's crazy is and there's so many little things like this in the story where you're like people onna, look back on this and think, wow, we were got down student at least I hope that. But OpenAI has two years from closed to convert, from a nonprofit to a four profit. All their funding will convert into debt.

I think it's six to nine percent interest. Hey, this isn't good. Also, at some point, open eye is gna have to work out a way to go public, like said, because otherwise, why did people invest? Why did people bother? They could do future secondary sales.

And at some point, if that's all they are going to be able to do in secondary sales, referring to selling private stock to another individual. I mean, at that point, it's upon on this game, you're just pulling in money to hand out money to other people eventually. It's not good.

This this company has no positive profitability. IT doesn't have one. And regardless of what happens with everything I mentioning, they still need to raise more funds.

ChatGPT free version is an actual poison on open A S system. It's a marketing channel that burns billions of dollars to introduced people to a product that only ten million people will actually pay for. An open your ice future depends large on its ability to continue convincing people to use IT.

So how does this continue? How does open a ee survive? I don't think I can. OpenAI is a disaster in the making and behind its its a nasta shea disaster.

A lack of fundamental strengths in the generate A I market rate, large, if open eye, can only make a billion dollars as the leader in this market and really I mean eight hundred million because the other two hundred million comes in. Microsoft IT suggests that there's not really develop or user interest in generate V I products at large. Perhaps it's the hallucination problem where authority timely states something that isn't true.

Or maybe it's just that generate A I isn't something that produces interesting interactions with the user. While you could argue that somebody could work out a really cool product, it's time to ask why amazon, google, meta, OpenAI, apple and microsoft have failed to make one one in the last two years. Where is IT? Where's the killer up? And no, it's not early days.

Shut up. More money than anything has ever raised is gone into these companies. More attention has gone into these companies than any other hyper scale movement ever.

Even the matter of us, where's the product, man? Where is IT? And while this bubble can continue coasting for a little while longer, nothing about the open a eyes story looks good.

His companies lost the bleeding money with every single interaction with the customer and flogged serve software that's best kind of useful in the actively harmful to the environment unless something significantly changes like a massive scientific breakthrough, an energy or computer efficiency. I just don't see how OpenAI e makes IT two more years. And we're still if my hypothesis about the wider market is true, there might just not be a viable business in making and selling large language models.

While matter in OpenAI might be able to claim hundreds of millions of users on these services. I don't see any evidence that these people will make them any money. And in fact, I only see evidence that lose them if i'm right.

We're watching VS and companies like microsoft, even tens of billions of dollars to power the next generation of products, and nobody really gives a shit of out. And in the next episode, i'm going to go into the much bigger problem, the sub prime ei crisis, where everyone building on these platforms is inevitably gonna a get rug pod and open a anthropic. And other companies raised their Prices because if they don't raise their Prices, how are they going to be able to afford to keep going? I don't bloody now, but get into the next episode.

Thank you for listening to Better roof line. The editorial um composer of the battle offline theme song is matter salesa pe. You can check out more of his music and audio projects.

A meta salty dot com M A T T O S O W S K I dot com, you can email me, are easy at Better off line dot com, or visit Better off line 到 com to find more podcast links。 And of course, my newsletter, I also really recommend you go to chat, that was your ed or ant to visit the discord and go to ask lash Better roof line to check out our red IT. Thank you so much for listening. Better off line is a production of cool zone media. For more from coal zone media, visit our website calls on media dot com, or check us out on the I heart radio APP apple podcast or whatever you get your podcast.

Did you know your life insurance is an asset you can sell? M, J. Jack and C, E, O of advocates. For almost twenty years, advocates has been purchasing life insurance policies from seniors just like you and IT advocates pays that com.

We've created a free policy value calculator so you can find out the worth of your policy in just second, there are no fees and no obligations. Get real value for your life insurance when you need IT with advocates. Learn more at advocates.

Pays that com. 毛豆。

M O T O I U S for the best replay social caso experience wherever you are, model offers a huge selection of bigger style games with free spins, exciting promotions and always jack. But you can waste your time with the others or you can win register today at M O G O U S for your .

free welcome ones. Moto a accel conditions of website put to. Once upon a time, amazon sic deo books, listeners everywhere rejoice.

Now listen to the from an enormous library of popular audio book titles, including romantic acy autobiography, true crime. And more suddenly, listener's didn't mind sitting in traffic or even missing their flight. Amazon music limited now includes auto download, the amazon music APP now to start listening, terms apply.

Whether you're ordering wins for the game, wiping up a seven later did or ordering pizza, there's something about football that makes you want to eat. And this football season, uber is has the best deals on game day food, no matter what you're craving from two for one pizza to buy one, get one wings. Uber its will be dropping new deals each week all season long. Ubs official on demand delivery partner of the nf water now terms and conditions apply, see apr details.

The best vacation stories happen when everything feels effortless. As america's favorite vacation company for fifty five years, booking with applications is like turning on easy mode. You can find the best deals all on one side and enjoy added per. So all that love to do is let the stories unfold, turn on easy mode, and start your story today with apple vacations, visit apple vacations stop com, or call your local travel advisor.