OpenAI Cannot Survive

This is an iHeart podcast.

Hi, I'm Morgan Sung, host of Close All Tabs from KQED, where every week we reveal how the online world collides with everyday life.

There was the six-foot cartoon otter who came out from behind a curtain.

It actually really matters that driverless cars are going to mess up in ways that humans wouldn't.

Should I be telling this thing all about my love life?

I think we will see a Twitch stream or a president maybe within our lifetimes.

You can find Close All tabs wherever you listen to podcasts.

There's more to San Francisco with the Chronicle.

There's more food for thought, more thought for food.

There's more data insights to help with those day-to-day choices.

There's more to the weather than whether it's going to rain.

And with our arts and entertainment coverage, you won't just get out more, you'll get more out of it.

At the Chronicle, knowing more about San Francisco is our passion.

Discover more at sfchronicle.com.

Who knew you could get all your favorite summer fruits and veggies in as fast as an hour with Walmart Express delivery?

Crisp peppers, juicy peaches, crunchy cucumbers, and more at the same low prices you'd find in store.

And freshness is guaranteed.

If you don't love our produce, contact us for a full refund.

You're definitely going to need a bigger salad bowl.

Order now in the app.

The Walmart you thought you knew is now new.

Subject to availability, fees, and restrictions apply.

Adobe Acrobat Studio, so brand new.

Show me all the things PDFs can do.

Do your work with ease and speed.

PDF spaces is all you need.

Do hours of research in an instant.

With key insights from an AI assistant.

Pick a template with a click.

Now your prezo looks super slick.

Close that deal, yeah, you won.

Do that, doing that, did that, done.

Now you can do that, do that, with Acrobat.

Now you can do that, do that, with the all-new Acrobat.

It's time to do your best work with the all-new Adobe Acrobat Studio.

CoolZone Media.

Hello and welcome to Better Offline.

I'm your host, Edzy Tron.

Now, before we go any further, I hate to ask you to do this, but I need your help.

I'm up for this year's Webbys for the the Best Business Podcast Award.

I know it's a pain in the ass to register but can you sign up and vote for Better Offline?

Never won an award in my life and I have enough listeners that I think we can tip the scales.

But on with the show.

The following two-part episode is the culmination of months of research presenting a case I've been making in part since July of last year.

Open AI is a financial abomination, a thing that should not be, an aberration, a symbol of rot at the heart of Silicon Valley, a company that unrepentantly and needlessly burns billions of dollars with no end in sight, helmed by a critinous, critinous, disingenuous billionaire who continually lies about what it is that it will do because he, like I've been saying since last year, knows that generative AI can't do much more than it does today.

What I'm going to lay out for you is my case that OpenAI can't survive, that it's borderline impossible under any of the current terms for this company to continue and that its demise will bring about the death of or significant harm of several other firms.

A cottage industry of despicable billion-dollar burn rate capitalistic monstrosities has sprung up around this stupid fucking company in the hopes of further inflating a bubble set to burst at any moment.

My disgust for the parties involved is unrelenting, as is my disgust for a media industry that failed to even attempt to tackle the subject matter I'm going to detail.

I believe at the end of these episodes you'll see my point, and at the very least agree that OpenAI's current situation is totally untenable.

If I'm right, OpenAI will go down in history as an abdication of due diligence, fiscal responsibility, and frankly common sense, both in the venture capitalists and entities that propped it up and a tech media that was more concerned with taking detailed notes on my comings and goings than knowing their arse from their earhole.

Those who have failed to hold men like Sam Altman and Dara Amadei accountable have tacitly approved a financial and environmentally destructive movement that will lead to very little actually happening or changing in the world other than damaging our power grid and the theft of art from millions of people.

These episodes will be full of numbers and statements and very few declarations or personal opinions, though probably a few swear words if I'm honest.

I don't need to get that personal here.

The numbers in question, they're damaging, they're staggering, they're worrying and ultimately spell collapse.

The truth is we don't even need to talk about tariffs for things to go sideways for this industry.

The price of a GPU could rise 100% or it could halve.

It really wouldn't make much of a difference to OpenAI's chances of survival.

That's how bad the fundamentals are.

And to illustrate that point, I'm going to ask a number of relatively simple questions over the next couple of episodes and make an attempt to answer them.

First, let's start with something simple.

How much cash does OpenAI have?

At the start of April, OpenAI closed what was called the largest private tech funding round in history, where it raised an astonishing $40 billion.

And the reason I'm saying this with a sarcastic inflection is that OpenAI has only actually raised $10 billion of the $40 billion, with the rest arriving by the end of the year.

And even as I record this, I don't know if the money's actually gone yet.

I'll get into that in a minute.

Now, a lot can happen in a year, and the remaining $30 billion, $20 billion of which will allegedly be provided by SoftBank, is partially contingent on OpenAI's conversion from a non-profit to a for-profit by the end of 2025.

And if it fails, SoftBank will only give OpenAI a further $20 billion.

I'll get into how fucking stupid this gets later.

The round also valued OpenAI at an astonishing $300 billion.

To put that in context, OpenAI had revenues of $4 billion in 2024.

This deal values OpenAI at 75 times its revenue.

That's more than Tesla even at its most ludicrous peak.

I also want to add that, as of writing this sentence, this money is yet to arrive.

Maybe it will arrive by the time this is out.

Maybe it won't really make me look stupid.

But based on SoftBank's filings, which I'll link to in the spreadsheet for this episode, say that the money will arrive mid-April, it's April 14th as I'm recording this, and that SoftBank would be borrowing as much as $10 billion of the financing for the round, with the option option to syndicate, meaning you bring in other investors, the rest of it.

For the sake of argument, I'm going to assume the money actually arrives, though.

Filings also suggest that, and I'm quoting, in certain circumstances, the second $30 billion tranche could arrive in early 2026.

This isn't great.

It also seems that SoftBank's $10 billion commitment is contingent on getting a loan, as it says it's financed through borrowings.

from Mizuho Bank Limited among other financial institutions.

SoftBanks had plenty of loans in the past, so I think they're going to get it, but I think this is one of their biggest.

OpenAI also revealed it now has 20 million paying subscribers and over 500 million weekly active users.

If you're wondering why it doesn't talk about monthly users, it's because they're likely much higher than 500 million, which would reveal exactly how poorly OpenAI converts free chat GPT users to paying ones.

The information reported back in January that OpenAI was generating $25 million in revenue a month from its $200 a month pro subscribers.

And just so we're clear, they lose money on every one of those too.

Suggesting that they have around 125,000 ChatGPT pro subscribers, each losing the money somehow.

Assuming the other 19,875,000 users are paying $20 a month, that puts revenue at about $423 million a month, or at least $5 billion a year from ChatGPT subscriptions.

This is what reporters mean when they say annualized revenue, by the way.

It's literally the month, the monthly revenue, the money they're making in one month multiplied by 12.

And you'll be surprised to hear that people play silly buggers with that all the time.

In March, Bloomberg reported that OpenAI expects its revenue to triple to $12.7 billion in 2025.

Assuming a similar split of revenue to 2024, this would require OpenAI to nearly double its annualized subscription revenue from Q1 2025, from $5 billion right now to around $9.27 billion, and nearly quadruple API revenue from 2024's revenue of $1 billion, which includes Microsoft's 20% payment for access to OpenAI's models.

And that would get them about 3.43 billion.

You don't need to super worry about these numbers.

And I realize these are messy numbers.

It's just unclear how OpenAI intends to pull any of this off.

It's an incredible leap, and OpenAI's own plans don't exactly inspire confidence.

They're really good at getting free subscribers.

They don't seem to be able to get paying ones in quite the same number.

And also, they even lose the money.

Every time I think about this company, I start feeling a little crazy if I'm on it.

Anyway, the information reported in February that the OpenAI planned to grow its revenue by making $3 billion a year selling agents, with ChatGPT subscriptions $7.9 billion and API costs $1.8 billion making up the rest.

This is, of course,

what's the technical term?

Oh, it's bollocks.

It's complete fucking bollocks.

I'm sorry.

Agents, by the way, are AI chat bots that can do something, like interact with another program on the user's behalf.

OpenAI's agents can't even do the simplest tasks, and $3 billion of the $12.7 billion figure appears to be a commitment made by SoftBank to purchase $3 billion dollars a year of open ai's tech now let's pass out these numbers precisely so incoming monthly revenue roughly 425 million dollars give or take theoretical revenue from softbank's completely made up thing 250 million dollars a month however i really can find no proof that softbank has begun to make these payments or indeed that it intends to make them even how it intends to now let's talk liquidity OpenAI has $10 billion that they're yet to receive ever sort of recording this, but let's assume they get it.

And And it will be $10 billion, $7.5 billion from SoftBank and a syndicate of investors, including Microsoft and Cartoon and a few others, potentially.

There's also an indeterminate amount of remaining capital on the $4 billion credit facility provided by multiple banks back in October 2024, and that was raised alongside a funding round that valued the company at $157 billion.

Now, as a note, the October announcement stated that OpenAI had access to over $10 billion in liquidity, giving us a sense of how fast it's burning through the cash it has on hand, as that was pretty much all the money they'd raised at the time.

Based on reports, OpenAI will not have access to the rest of the $40 billion that SoftBank is funding them with until the end of the year, and it's unclear what part of the end of the year that is, but SoftBank's filings says December.

Will it be December 1st?

Will it be Christmas time?

Will Sam Altman look onto the tree and Masayoshi-son will be there?

We'll find out, won't we?

But we can assume in this case that OpenAI likely has, in the best-case scenario, access to about roughly $16 billion in liquidity at any given time.

It's reasonable to believe that OpenAI will raise more debt this year, despite this massive raise, and I'd estimate it does so to the tune of around $5 or $6 billion.

Without it, I'm genuinely not sure what they're going to do.

And as a reminder, kids, OpenAI loses money on every single user, free or paying.

Be honest, how many tabs do you have open right now?

Too many?

Sounds like you need close All Tabs from KQED, where I, Morgan Sung, Doom Scroll so you don't have to.

Every week, we scour the internet to bring you deep dives that explain how the digital world connects and divides us all.

Everyone's cooped up in their house.

I will talk to this robot.

If you're a truly engaged activist, the government already has data on you.

Driverless cars are going to mess up in ways that humans wouldn't.

Listen to Close All Tabs wherever you get your podcasts.

Let's be real.

Life happens.

Kids spill.

Pets shed.

And accidents are inevitable.

Find a sofa that can keep up at washable sofas.com.

Starting at just $699, our sofas are fully machine washable inside and out.

So you can say goodbye to stains and hello to worry-free living.

Made with liquid and stain-resistant fabrics.

They're kid-proof, pet-friendly, and built for everyday life.

Plus, changeable fabric covers let you refresh your sofa whenever you want.

Neat flexibility?

Our modular design lets you rearrange your sofa anytime to fit your space, whether it's a growing family room or a cozy apartment.

Plus, they're earth-friendly and trusted by over 200,000 happy customers.

It's time to upgrade to a stress-free, mess-proof sofa.

Visit washable sofas.com today and save.

That's washablesofas.com.

Offers are subject to change and certain restrictions may apply.

In business, they say you can have better, cheaper, or faster, but you only get to pick two.

What if you could have all three at the same time?

That's exactly what Kohir, Thomson Reuters, and specialized bikes have since they upgraded to the next generation of the cloud.

Oracle Cloud Infrastructure.

OCI is the blazing fast platform for your infrastructure, database, application development, and AI needs, where you can run any workload in a high-availability, consistently high-performance environment and spend less than you would with other clouds.

How is it faster?

OCI's block storage gives you more operations per second.

Cheaper?

OCI costs up to 50% less for computing, 70% less for storage, and 80% less for networking.

Better?

In test after test, OCI customers report lower latency and higher bandwidth versus other clouds.

This is the cloud built for AI and all your biggest workloads.

Right now, with zero commitment, try OCI for free.

Head to oracle.com slash strategic.

That's oracle.com slash strategic.

I couldn't even believe it was real.

Join me, Tatiana Siegel, executive editor of film and media at Variety, for a four-part tale of youthful ambition, artistic integrity, and the dark side of fame.

Just like my parents talk about they knew where they were when John F.

Kennedy was killed.

Pretty much everyone I know knows exactly where they were when River died.

Featuring new interviews with Samantha Mathis, Dr.

Drew Pinski, Corey Feldman, and more.

Listen to Variety Confidential on the iHeartRadio app or wherever you get your podcasts.

Now, here's my next question.

What are OpenAI's obligations?

Now, when I put out How Does Open AI Survive and Open AI is a Bad Business, I used reported information to explain how this company was, at its core, unsustainable.

I'll link to the newsletters in the spreadsheet and the podcast too, but let's refresh our memory, shall we?

Okay, compute costs.

At least $13 billion of which are going to Microsoft alone in 2025 and as much as $594 million to my favorite company, Core Weave.

Coreweave's back.

Every time I think I got away from Core Weave, they pop out like a sour turd coming back in the toilet.

Anyway, it seems from even a cursory glance that OpenAI's costs are increasing dramatically.

The information reported earlier in the year that OpenAI projects to spend $13 billion on compute with Microsoft alone in 2025, nearly tripling what it spent in total compute in 2024, which was $5 billion with $3 billion for training and $2 billion for running their models.

This suggests that OpenAI's costs are skyrocketing.

And that was before the launch of their new image generator, which led to multiple complaints from Sam Altman about a lack of available GPUs, leading to OpenAI's terrible little man CEO saying that he expected stuff to break and delays in new products.

Nevertheless, even if we assume OpenAI factored in the compute increases into its projections, it still expects to pay that $13 billion to Microsoft alone.

This number, however, doesn't include the $12.9 billion five-year-long compute deal signed with Coreweave, a deal that was a result of Microsoft declining to pick up the option to buy the compute themselves for OpenAI.

As an aside, from what I am hearing, and this is sources telling me, most of the Microsoft compute was OpenAI.

It was basically they put an OpenAI sticker over Microsoft.

It was basically the same thing.

Now, payments for this deal with OpenAI and Coreweave, they start in October of 2025, according to the information.

And assuming that it's even paid, or like evenly paid, or even like anyone gets any cash, but these contracts are weird.

You never know exactly how they're paid.

This would still amount to roughly $2.38 billion a year.

However, for the sake of argument, let's consider the payments are around $198 million a month.

Though there are scenarios, such as, say, Core Weave's build-out partner not being able to build the data centers or Coreweave not having the money to pay to build them, where OpenAI might end up paying less.

And don't worry, I'll get to that later.

Let's talk about Stargate.

You heard this?

You seen this?

Stargate is this data center project that OpenAI is going into.

And they've dedicated somewhere in the region of $19 billion to this project, along with another $19 billion provided by Software Bank at some indeterminate time.

And there are other partners too, like Oracle.

It's not even obvious what they're putting in either.

One thing I can tell you, though, is the Trump administration is not actually spending any money on this.

Anyone, I've had tons of of people emailing me being like, hey, Ed, Donald Trump was there.

Yeah, Donald Trump was there.

Doesn't mean shit.

Anyway, based on reporting from Bloomberg, OpenAI plans to have 64,000 Blackwell GPUs running by the end of 2026, or roughly about $3.84 billion worth of them.

I should also note that Bloomberg said that 16,000 of these chips would be operational by summer 2025, though it's unclear if that will actually happen.

And as I'll get to later, I got some more questions about what exactly is happening at Stargate, all told.

Though it's unclear who actually pays what parts of Stargate, it's safe to assume that OpenAI will have to at least put a billion dollars into a project that is meant to be up and running by the end of 2026, if not more.

As of now, Stargate really only has one like data center project under development.

It's in Abilene, Texas.

And as I've mentioned, it's not really clear how it's going.

Though a recent piece from the information reported that it was currently empty and incomplete and if it stays that way, and I quote again, OpenAI could walk away from the deal, which would cost Oracle billions of dollars.

Though the article takes great pains to assure the reader that that won't be likely, even an inkling of such a possibility is a bad fucking sign.

Business Insider's reporting on the site in Abilene calls it a $3.4 billion data center development, as did the press release from the developer Crusoe, who we'll get to later, though these numbers don't include GPUs, hardware, or the labor necessary to run them.

Right now, Crusoe is, according to Business Insider, building six new data centers, each with a minimum square footage, which will join the two it is already constructing for Oracle.

Oracle has signed, according to the information, a 15-year-long lease with Cruiser for its data centers, all of which will be rented to OpenAI.

In any case, OpenAI's exposure could be much, much, much higher than the billion dollars.

And I'll explain in greater depth how I reached the figure when I get there.

But nevertheless, it could be much higher.

If OpenAI has to contribute significantly to the costs of associating with Stargate in general, it could cost them a lot of money.

Data centers aren't something you can do funny money math with.

Data Center Dynamics reports that the Abilene site is meant to have 200 megawatts of compute capacity in the first half of 2025 and then as much as 1.2 gigawatts by mid-2026.

To give you a sense of the total cost for the project, former Microsoft VP of Energy Brian Janus said in January that it cost $25 million a megawatt, or about $25 billion a gigawatt, meaning that the initial capital expenses for Stargate to spin up its first 200 megawatt data center will be around $5 billion,

spiraling to $30 billion or more for the entire project.

The information is reported that the site, which could be, and I quote, potentially one of the biggest AI data centers, could cost $50 billion to $100 billion in the coming years.

Where's that fucking money coming from?

Assuming we stick with the lower end of these cost estimates, it's likely that OpenAI is on the hook for over $5 billion for the Abilene site based on the $19 billion it's committed to the overall Stargate data center project.

This expenditure won't come all at once and will be spread across several years.

Still, assuming even the rosiest numbers, it's hard to see how OpenAI doesn't have to pony up at least a billy in 2025.

And that's likely because the development of this site is going to be heavily delayed by both tariffs, labor shortages, and oracles, as reported by information.

Well, they're trusting, and this is the kind of quote you really want to hear.

They're trusting scrappy but unproven startups to develop the project.

Is that good?

Is that who you want doing this?

I know when I get a contractor in to fix something, my first thought is, is this guy scrappy?

Anyway, let's talk about the other costs.

They're at least 3.5 billion.

Based on reporting from the information from last year, OpenAI will spend at least $2.5 billion across salaries, data, referring to buying data from other companies, hosting, and other costs of sales, and sales and marketing, and then another billion on what infrastructure OpenAI owns.

I expect the latter costs to balloon with OpenAI's investment in physical infrastructure for Stargame.

There's another bloody question.

How does OpenAI meet its obligations?

Based on previous estimates, OpenAI spends about $2.25 to make a buck.

At that rate, it's likely that OpenAI's costs in its most rosy red revenue projections of $12.7 billion are at least $28 billion, meaning that it's on course to burn at least $14 billion in 2025 alone.

Assuming that OpenAI has

literally all of the money they have from last year, they don't, but for the sake of argument, let's pretend they have $10 billion, as well as $10 billion from SoftBank.

It's still unclear how they pay for everything.

Now, while OpenAI likely has preferential payment structures with all of their vendors, such as discounted rates with Microsoft or Azure Cloud Services, it will eventually have to pay someone, especially in the case of costs related to Stargate, many of which are upfront and involve physical things happening.

In the event that its costs are as severe as reporting suggests, OpenAI's revenue comes at a terrible cost and will likely be immediately funneled directly into funding the obscene costs behind inference and training models like GPT 4.5, which Sam Alton called a giant expensive model.

And yeah, nevertheless, he he pushed that to every single user.

Worse still, OpenAI has, while delaying its next model, GPT-5, promised to launch, after all, its O3 reasoning model after saying it wouldn't do so, which is strange because it turns out O3 is actually way more expensive to run than people thought.

With ArcPrice Foundation, a non-profit that makes the Arc AGI tests for benchmarking models, because real tests don't work on them, they need to make something out.

Well, they're estimating that it will cost about 30 grand a task, which is that's a lot of money.

And yes, they, I get it, OpenAI has a new sugar daddy in the form of SoftBank, but SoftBank has to borrow money to meet its obligations for Stargate and also OpenAI.

And this is leading to its financial condition likely deteriorating.

And that's S ⁇ P Global.

You know, the S ⁇ P 5, those people, those people said that.

That's not what you want to hear.

Let's talk SoftBank.

As of right now, SoftBank has committed to the following.

At least $30 billion in funding as part of OpenAI's recent $40 billion funding round.

Now, SoftBank's filings surrounding OpenAI's funding also suggest that they are ultimately on the hook for the entire $40 billion, but they can syndicate it, like I mentioned earlier.

Reporting suggests that the syndication will happen, and it will happen with people like Coutu, Microsoft, and other investors.

Now, here's the funny part.

If OpenAI fails to convert to a for-profit, that $40 billion is slashed down to a paltry $30 billion.

Although, again, SoftBank's share is contingent upon whether it finds other investors to join the deal.

Now, there's another 3 billion dollars that SoftBanks promised to spend on OpenAI's tech, and then $19 billion for the Stargate data center project, which SoftBank is taking full financial responsibility for, the information reports.

And the total is either $52 billion or $62 billion, with at least $10 billion due by the end of 2025, but more like 20, like there's so much money.

To be clear, SoftBank had to borrow all of the $10 billion.

How the fuck are they meant to get the 30 billion dollars now i kind of mentioned s p but i want to really return to that now softbank's exposure to open ai is materially harming the company and i'm quoting the wall street journal here ratings agency s p global said tuesday that softbank's financial condition will likely deteriorate as a result of the open ai investment and that its plans to add debt could lead the agency to consider downgrading softbank's ratings While one might argue that SoftBank has a good amount of cash, the journal also adds that they're somewhat hamstrung in its use as a result of CEO Masayoshi's son's dick shit, reckless gambles.

Again, quoting from the Wall Street Journal, SoftBank had a decent buffer of $31 billion of cash as of December 31st, but the company has also pledged to hold much of that in reserve to quell worried investors.

SoftBank has committed to not borrow more than 25% of the value of its holdings, which means it will likely need to sell some of the other partners of its empire to pay the rest of the OpenAI deal.

Worse still, it seems that, as mentioned before, that SoftBank will be financing the entirety of the first 10 billion, 7.5 billion if they're able to syndicate it.

As a result, SoftBank is likely going to have to sell off parts of their actually valuable holdings in companies like Alibaba or Arm, or worse still, parts of their ailing investments from a Vision Fund, resulting in a material loss in its underwater deals.

This is an untenable strategy, and I'd like to explain why.

First, OpenAI needs at least $40 billion a year to survive, and its costs are only increasing.

While we don't have much transparency into OpenAI's actual day-to-day finances, we can make educated guesses that its costs are increasing based on the amount of capital it's raising.

If OpenAI's costs were flat or only mildly increasing, we'd expect to see raises roughly the same size as the previous ones, 6.6 billion, something in that range.

A $40 billion raise is nearly six times the previous funding round.

Admittedly, multiples like that aren't particularly unusual.

If a company raises 300 grand in a pre-seed round and a $3 million Series A round in funding, that's a tenfold increase.

But we're not talking about hundreds of thousands of dollars or even millions of dollars we're talking about billions of dollars if open ai's funding round with softbank goes as planned it will raise the equivalent of the entire gdp of estonia a fairly wealthy country itself that's also a member of nato and the european union that alone should give you a sense of how fucking stupid this all is

Stupid, sure, but undoubtedly necessary.

Per the information OpenAI plans to spend as much as $28 billion in compute on Microsoft Azure Cloud in 2028, over a third of OpenAI's revenue per the same article will come from SoftBank's alleged spend.

It's reasonable to believe that OpenAI will, as a result, need to raise in excess of $40 billion a year, though it's reasonable to believe that it will need to raise more like $50 billion a year until they reach profitability.

You know, never.

And this actually has a reason.

It's due to the growing cost of doing business as well as the various infrastructure commitments they've made, both in the terms of Stargate as well as deals with third-party suppliers like Coreweave and indeed Microsoft.

OpenAI CEO Sam Altman's statement around costs also suggests that they're going up quite fiercely.

In late February, he claimed that OpenAI was out of GPUs.

While this suggests that there is demand for some products, like its horrifying image-generating tech that has made abominations that insult Miyazaki, and I will hate them for it forever, and thus made them go viral in March, it also means that to meet the demand of the horrible abominations, OpenAI needs to spend more.

And at the risk of repeating repeating myself, that demand doesn't necessarily translate into revenue or profitability.

Now, second, SoftBank cannot fund OpenAI long term.

I must be clear, OpenAI's costs are projected to be $320 billion over the next five years.

SoftBank has to overcome significant challenges to fund both OpenAI and its part of Stargate.

And when I say fund, I mean fund the current state-of-both projects, assuming no further obligations or complications.

The information reports the OpenAI forecast that it will spend, as I mentioned, $28 billion on compute with Microsoft alone in 2028.

And I apologize for repeating myself, these numbers fucking matter.

The same article also reports that OpenAI would turn profitable by the end of the decade after building out Stargate, suggesting that OpenAI's operating expenditures will grow exponentially year over year.

According to the information, it expects its costs to surpass $320 billion between 2025 and 2030, with half of that going towards funding model training and development.

How the fuck does building Stargate make them profitable?

I really can't, anyway.

It won't, it's the same shit.

But SoftBank has had to, and will continue having to, go to remarkable lengths to fund OpenAI's current $40 billion round, lengths so significant that it may lead to their credit rating being further downgraded, as I mentioned.

Even if we assume the best case scenario, OpenAI successfully converts to a for-profit entity by the end of 2025 and gets that full $30 billion,

it seems unlikely, if not impossible, for it to to continue raising the amount of capital they need to to continue operations.

As I've argued in multiple newsletters and podcasts, there are only a few entities that can provide the kind of funding that OpenAI needs.

These include big tech-focused investment firms like SoftBank, sovereign wealth funds like those of Saudi Arabia and the UAE, and perhaps the largest tech companies.

I also want to be clear that I keep getting messages being like, the government could do it, Baylor.

Couldn't it?

What if the government did it?

It's not going to happen.

Stop fucking asking me.

It's just not going to happen.

The government's not sending $40 billion to Sam Altman.

They're not going to do it.

I will apologize personally to each and every one of you if I'm wrong, but they will need to send $40 billion cash to OpenAI.

But these entities can meet OpenAI's needs for some time, but not all the time.

It's not...

realistic to expect SoftBank or Microsoft or the Sauds or Oracle or whoever to provide $40 billion every year for the foreseeable future.

And just to be clear, this is what OpenAI needs.

Eventually, even the Sauds have to have a break.

And I don't know if you remember from a previous episode, but Masayoshi-san gets a lot of his money from them.

And they're not happy with him.

In fact, Masayoshi-san said a few years ago that he owed Mohammed bin Salman.

Not a great guy to owe?

Anyway, this really is especially true for SoftBank, by the way.

They're bruised, they're battered after several rough years, including a failed multi-billion dollar investment in WeWork.

Based on its current promise to not borrow more than 25% of its holdings, it's near impossible that SoftBank will be able to continue to fund OpenAI at this rate.

And $40 billion a year may not actually be enough.

Based on SoftBank's last reported equity value of its holdings, they have about $229 billion of stuff, meaning that they can borrow just over $57 billion while remaining compliant with these guidelines.

In any case, it's unclear how SoftBank can fund OpenAI, but it's far clearer that nobody else is willing to.

Be honest, how many tabs do you have open right now?

Too many?

Sounds like you need Close All Tabs from KQED, where I, Morgan Sung, Doom Scroll so you don't have to.

Every week, we scour the internet to bring you deep dives that explain how the digital world connects and divides us all.

Everyone's cooped up in their house.

I will talk to this robot.

If you're a truly engaged activist, the government already has data on you.

Driverless cars are going to mess up in ways that humans wouldn't.

Listen to Close All Tabs, Tabs, wherever you get your podcasts.

Tired of spills and stains on your sofa?

WashableSofas.com has your back, featuring the Annibay Collection, the only designer sofa that's machine-washable inside and out, where designer quality meets budget-friendly prices.

That's right, sofas started just $699.

Enjoy a no-risk experience with pet-friendly, stain-resistant, and changeable slip covers made with performance fabrics.

Experience cloud-like comfort with high-resilience foam that's hypoallergenic and never needs fluffing.

The sturdy steel frame ensures longevity and the modular pieces can be rearranged anytime.

Check out washable sofas.com and get up to 60% off your Anibay sofa, backed by a 30-day satisfaction guarantee.

If you're not absolutely in love, send it back for a full refund.

No return shipping or restocking fees.

Every penny back.

Upgrade now at washablesofas.com.

Offers are subject to change and certain restrictions may apply.

I couldn't even believe it was real.

Join me, Tatiana Siegel, executive editor of film and media at Variety, for a four-part tale of youthful ambition, artistic integrity, and the dark side of fame.

Just like my parents talk about they knew where they were when John F.

Kennedy was killed.

Pretty much everyone I know knows exactly where they were when River died.

Featuring new interviews with Samantha Mathis, Dr.

Drew Pinski, Corey Feldman, and more.

Listen to Variety Confidential on the iHeartRadio app or wherever you get your podcasts.

Every business has an ambition.

PayPal Open is the platform designed to help you grow into yours with business loans so you can expand and access to hundreds of millions of PayPal customers worldwide.

And your customers can pay all the ways they want with PayPal, Venmo, Pay Later, and all major cards so you can focus on scaling up.

When it's time to get growing, there's one platform for all business: PayPal Open.

Grow today at paypalopen.com.

Loan subject to approval in available locations.

Now, we're going to move off the questions for a minute because I just want to get into some problems I've popped.

Because it turns out that OpenAI, they got doo-doo-ass.

That's a professional finance term.

I did just say finance, and we're just going to keep that.

Now, OpenAI has started running into capacity issues, and this is a real problem.

And it suggests material instability in their business or infrastructure.

And it's really not clear clear how OpenAI expands further.

Let me explain.

It's important to note that OpenAI does not really have any of its own compute infrastructure.

The majority of its compute is provided by Microsoft though as mentioned previously OpenAI now has a deal with Coreweave to take over the capacity that Microsoft was going to have $12 billion or so of capacity in the future.

Anyway, in the last 90 days Sam Altman has complained about the lack of GPUs and pressure on OpenAI's servers many times.

In my newsletter published a few days ago, I'll listed six such examples should you be curious.

These statements in a bubble seem either harmless or like OpenAI's growth is skyrocketing, the latter of which might indeed be true but bodes ill for a company that burns money with every single user.

Any mention of rate limits or performance issues suggests that OpenAI is having significant capacity issues and at this point it's unclear what further capacity it can actually expand to outside of that currently available.

Samuel was complaining about melting GPUs.

You've got the lead from Sora saying, yeah, we're going to have some problems showing you stuff for a minute, Just because, like, yeah, we're melting GPUs.

There was a whole thing about how you should expect delays on product launches and service problems.

None of it's really good.

And like I said, it isn't really obvious how OpenAI is going to expand much further.

Remember, Microsoft has now pulled as much as two gigawatts of data center projects, walked away from a billion-dollar data center development in Ohio and declined the option, the one I just mentioned, on $12 billion of compute from CoreWeave that OpenAI had to pick up.

Meaning that OpenAI may be pushing up against the limits of what is physically available.

While the total available capacity of GPUs at many providers like Lambda and Crusoe is unknown, and indeed I don't know if Crusoe has a single data center at this point, we know that Core Weave has approximately 360 megawatts available compared to Microsoft's 6.5 to 7.5 gigawatts, a large chunk of which I think powers OpenAI.

If OpenAI is running into capacity issues, it could be one of the following.

They could be running up against the limit of what Microsoft has available or is willing to offer the company the information reported in october 2024 that open ai was frustrated with microsoft and said it wasn't moving fast enough to supply open ai with servers and now it could also be that while open ai's capacity is sufficient it does not have the resources available to easily handle bursts in user growth in a stable manner Per the information's reporting, Microsoft promised OpenAI 300,000 NVIDIA GB200 Blackwell chips by the end of 2025, or roughly $18 billion worth of them.

It's unclear if this has changed since Microsoft allowed OpenAI to seek other compute from other companies in late January 2025.

I also don't believe that OpenAI has any other viable options for existing compute infrastructure outside of Microsoft.

CoreWeab's current data centers mostly feature NVIDIA's aging Hopper GPUs, and while it could, and likely is, retrofitting its current infrastructure with Blackwell chips, doing so isn't easy or cheap.

Blackwell chips require far more powerful cooling and server infrastructure to make them run smoothly, a problem which led to a delay in their delivery to most customers, according to the information.

And even if Coreweave was able to replace every last hopper GPU with a black well, and they won't, it still wouldn't match what OpenAI needs to expand.

One might argue that it simply needs to wait for the construction of the Stargate data center or for Coreweave to finish the gigawatt or so of construction it's working on.

I want to be clear how impossible that is.

I need to be clear.

One of my least favorite responses to my work is people saying, they'll just build more data centers.

They'll just go and build them, man.

They're just going to build them right now.

They won't.

You can't just, data centers, don't just grow from the fucking ground.

As I've argued in the past, I have serious concerns over the viability of Coreweave ever completing its alleged contracted 1.3 gigawatts of capacity.

Based on calculations, it will have to spend in excess of 39 billion dollars to build it.

It's unclear how that'll happen, and it doesn't have the money to do so.

Like, it actually does not have the cash.

I'll get into that later, but they don't have the money.

However, even if I were to humor this idea, it's impossible that any of this project is done by the end of 2025.

And I'd argue, even in 2026, I can find no commitments to any time scale other than the fact that OpenAI will allegedly start paying Coreweave in October per the information, which could very well be using their current capacity.

I also can't find any evidence that Crusoe, the company building the Stargate data center in Texas, has any compute anywhere else.

Lambda, a GPU compute company that raised $320 million earlier in the year and according to Data Center Dynamics, operates out of co-location data centers in in San Francisco, California, and Allen, Texas, and is backed by more than 820 million in funds raised this year.

All of that just kind of doesn't say that they have a data center at all.

A co-location, that means you're in someone else's.

Buddy, you don't own a house, you're renting one.

And just to be clear, OpenAI's ability to scale is entirely contingent on the availability of whatever data center providers it has relationships with.

And all of the growth's coming from

these two companies.

I'll get there, don't worry.

But every time I kind of say this stuff out loud, I feel my soul slightly strip from me.

I feel like I'm in hell.

Because this is an insane thing.

If you or I went to someone and went like, hey, look, I'm going to lose $5 billion, but I promise you God will come out of the computer.

They'd institutionalize you or me.

They definitely would institutionalize me.

It's just fucking stupid.

But in any case, this means that OpenAI's only real choice for GPUs is either Coreweave or Microsoft.

And while it's hard to calculate precisely, OpenAI's best case scenario is that 16,000 GPUs come online in the summer of 2025 as part of the Stargate data center project.

And that's a drop in the bucket compared to the 300,000 of the fucking things that Microsoft have previously promised.

Now, one last thing.

Any capacity or expansion issues that happen with OpenAI will kneecap this company.

OpenAI is, regardless of how you or I may feel about generative AI, one of the fastest growing companies of all time.

It currently has, according to its own statements, over 500 million weekly active users.

Putting aside that each user is unprofitable, such remarkable growth, especially as it's partially as a result of its extremely resource-intensive image generator, is a massive, horrifying strain on their infrastructure.

The vast majority of OpenAI's users are free customers using ChatGPT, with only, like I mentioned earlier, 20 million paying subscribers and the vast majority of them on the cheapest 20 buck plan.

Open AI services, even in the case of image generation, are relatively commoditized, meaning that users can, if they really care, go and use any number of other large language model services.

They could use Bing, they could use Stable Diffusion, they could even use Grok if they really I don't like saying it.

But free users, they're a burden on the company, especially with such piss-poor conversion rates, losing it money with each prompt, which is, by the way, also case with paying subscribers, and the remarkable popularity of its horrible image generator.

It only threatens to bring more burdensome one-off customers that will generate a few abominable studio Ghibli pictures of Garfield with giant knockers and then never return.

If OpenAI's growth continues at this rate, it will run into capacity issues, and it does not have much room to expand.

While we don't know how much capacity they're taking up with Microsoft, or indeed whether Microsoft is approaching capacity or otherwise limiting them, we do know that OpenAI has seen reason to beg for access for more GPUs.

In simpler terms, even if OpenAI wasn't running out of money, even if OpenAI wasn't horrifyingly unprofitable, it also may not have enough GPUs to continue providing its services in a reliable manner.

If that's the case, there really isn't that much that can be done other than significantly limiting free users' activity on the platform, which is OpenAI's primary mechanism for revenue growth and customer acquisition, limiting activity or changing the economics behind its pay.

product.

And this is quoting Sam Orman, they could potentially find some way to let people pay for for compute they want to use more dynamically.

That's not good.

Altman's come up with some other ideas, like an idea for paid plans on March 4th, where 20 bucks a month goes to credits, which you can use across features like Deep Research, O1, GPT 4.5, Sora, and so on, with no fixed limits per feature.

And you choose what you want.

If you run out of credits, you can buy more.

I just want to be clear that that is a terrible fucking deal.

We have no idea what the credits would be.

And it would definitely be rigged so that you would have to buy more.

He's also brought up things like mentioning losing 200 bucks a month on the pro subscription.

But here's a funny one.

Buried in an article from the information from March 5th is a comment that suggests that OpenAI is considering measures like changing its pricing model entirely.

With Altman reportedly telling developers in London in February 2025 that OpenAI is primed to charge 20 or 30% of pro customers a higher price because of how many research queries they're doing.

But he suggested an a la carte or pay-as-you-go approach.

When it comes to agents though, we have to charge more than $200 a month.

Fucking just this fucking guy.

The problem is that with all of these measures, even if they succeeded in generating more money for the company, also need to reduce the burden on OpenAI's available infrastructure.

Remember, remember, data centers can take three to six years to build.

And even with Stargate's accelerated and I'd argue unrealistic timelines, OpenAI isn't unlocking a tenth of the promised compute.

that Microsoft gave them.

300,000 GPUs is a lot.

16,000?

Really not.

So what might these capacities look like?

What are the consequences?

You love payole horses.

I love payole horses.

Let's get riding.

Though downtime might be an obvious choice, capacity issues at OpenAI will likely manifest in hard limits on what free users can do, some of which I've documented previously.

Nevertheless, I believe the real payole horses of capacity issues come from arbitrary limits on any given user group, meaning both free users and paid users.

Some limits on what a user can do, a reduction in the number of generation of images for paid users, any any introduction of peak hours, or any increases in prices are a sign that OpenAI is running out of GPUs.

It's already said as happening publicly.

However, the really obvious thing, the real obvious pale horse would be service degradation, delays in generations of any kind, 500 status code errors, or chat GPT just failing to produce an answer.

OpenAI has, up until this point, had fairly impressive uptime.

Still, if it's running up against the wall, this streak will end.

The consequences depend on how often these issues occur and to whom they occur.

If free users face service degradation they'll bounce off the product as their use is likely far more fleeting than the paid user which will begin to erode OpenAI's growth.

Ironically rapid and especially unprecedented growth in one of OpenAI's competitors like XAI or Anthropic could also represent a palehorse for OpenAI though based on the monthly active users I've seen from Anthropic I don't think that's going to be a problem.

Now, if paid users face service degradation, it will likely cause the most harm to the company as while paid users still lose OpenAI money in the end, at least they receive some money.

Open AI is effectively one choice here.

Get more GPUs from Microsoft and

really its future depends heavily on both Microsoft's generosity and there being enough of them at a time when Microsoft has pulled back from over two gigawatts of data centers, specifically according to TD Cohen, because of it moving away from providing compute to OpenAI.

Now, admittedly, OpenAI has previously spent more on training models than inference, and that's the actual running of them.

And the company might be able to smooth down time time issues by shifting capacity.

This would, of course, have a knock-on effect on its ability to develop new models, and the company is already losing ground, particularly when it comes to Chinese rivals like DeepSeek.

Now, I know this has been a long episode, and the fact is, I'm not even close to finish.

I have some more tough, tough questions and tough problems for OpenAI.

Tune in to the next episode to hear them.

I realize this has been a lot, and I know you're very patient with me, you let me read my long scripts.

But this stuff is important.

It's important to me, and I think you're going to find it important too.

And I'll have a nice, sexy conclusion to this at the end of the next one.

But before I go, please vote for me in the fucking Webbies.

I want to win.

It's my pin post on Blue Sky and Twitter.

My handle's headzitron.com on BlueSky.

Please vote for me in the Webbies.

Anyway, catch you next episode.

Thank you for listening to Better Offline.

The editor and composer of the Better Offline theme song is Matosowski.

You You can check out more of his music and audio projects at matosowski.com.

M-A-T-T-O-S-O-W-S-K-I dot com.

You can email me at easy at betteroffline.com or visit betteroffline.com to find more podcast links and of course my newsletter.

I also really recommend you go to chat.where's your ed.at to visit the Discord and go to r slash betteroffline to check out our Reddit.

Thank you so much for listening.

Better Offline is a production of CoolZone Media.

For more from CoolZone Media, visit our website, coolzonemedia.com or check us out on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

Hi, I'm Morgan Sung, host of Close All Tabs from KQED, where every week we reveal how the online world collides with everyday life.

There was the six-foot cartoon otter who came out from behind a curtain.

It actually really matters that driverless cars are going to mess up in ways that humans wouldn't.

Should I be telling this thing all about my love life?

I think we will see a Twitch stream or a president, maybe within our lifetimes.

You can find Close All Tabs wherever you listen to podcasts.

Let's be real.

Life happens.

Kids spill.

Pets shed.

And accidents are inevitable.

Find a sofa that can keep up at washable sofas.com.

Starting at just $699, our sofas are fully machine washable inside and out.

So you can say goodbye to stains and hello to worry-free living.

Made with liquid and stain-resistant fabrics.

They're kid-proof, pet-friendly, and built for everyday life.

Plus, changeable fabric covers let you refresh your sofa whenever you want.

Neat flexibility?

Our modular design lets you rearrange your sofa anytime to fit your space, whether it's a growing family room or a cozy apartment.

Plus, they're earth-friendly and trusted by over 200,000 happy customers.

It's time to upgrade to a stress-free, mess-proof sofa.

Visit washablefas.com today and save.

That's washable sofas.com.

Offers are subject to change and certain restrictions may apply.

I couldn't even believe it was real.

Join me, Tatiana Siegel, executive editor of film and media at Variety, for a four-part tale of youthful ambition, artistic integrity, and the dark side of fame.

Just like my parents talk about they knew where they were when John F.

Kennedy was killed.

Pretty much everyone I know knows exactly where they were when River died.

Featuring new interviews with Samantha Mathis, Dr.

Drew Pinski, Corey Feldman, and more.

Listen to Variety Confidential on the iHeartRadio app or wherever you get your podcasts.

Ah, Smart Water Alkaline with Antioxidant.

Pure, crisp taste, perfectly refreshing.

Whoa, that is refreshing.

And a 9.5 plus pH.

For those who move, those who push further, those with a taste for taste.

Exactly.

I did take a spin class today after work.

Look at you.

Restoring like a pro.

I mean, I also sat down halfway through.

Eh, close enough.

Smart Water Alkaline with Antioxidant.

For those with a taste for taste, grab yours today.

This is an iHeart podcast.