The OpenAI Problem Is About To Become OUR Problem

0:00 Open AI has not been having a great year.

0:03 As it completes its metamorphosis from a non-profit

0:05 organization operating for the good of humanity

0:08 to a shareholder focused company gearing up

0:09 for one of the biggest IPOs in history,

0:11 it has come up against some uh teething problems.

0:16 It is facing legal challenges over the fair use of intellectual property,

0:19 legal challenges over its governance,

0:21 legal challenges over the safe use of its tools,

0:23 talent churn amongst its top developers, and now significant reputational damage

0:27 amongst effectively all of its stakeholders.

0:30 All of this wasn't helped by its willingness

0:32 to embrace what many see as war profiteering,

0:34 further straying away from their original stated goal

0:37 in order to chase revenue wherever it can find it.

0:39 Because, well, I mean,

0:41 fully autonomous AI kill bots and mass surveillance systems don't

0:44 really sound like they are improving the future of humanity.

0:47 In response to this, over just the last weeks,

0:49 some of even the most staunch AI optimists have

0:52 spoken out against the decisions this company has been making.

0:55 But, anyway, all of this is on top

0:57 of the biggest and most immediate problem of all,

0:59 which is that Open AI is simply burning

1:02 billions of dollars every month on top of trillions

1:04 of dollars in future spending commitments with no clear

1:07 answer for how it's going to get that money.

1:09 Now, maybe this would all be okay if it

1:12 was still the undisputed leader in this world-changing technology.

1:15 But, in the last 3 years,

1:17 several other models from better funded companies have come

1:20 along and either matched or exceeded Open AI's capabilities.

1:23 It's now unclear if the early pioneers

1:25 of this technology have spent hundreds of billions

1:27 of dollars developing systems that can now be

1:30 effectively replicated by random Swedish men in their bedroom.

1:33 It's also becoming less clear exactly how world-changing this technology

1:36 really is going to be in the first place,

1:38 putting more question marks over the company's future.

1:41 For now, its investors are doubling down.

1:43 Just last week, Amazon, who is theoretically a competitor, Nvidia, a supplier,

1:48 and the ever-reliable SoftBank announced a record

1:50 investment of $110 billion in additional funding,

1:53 which will keep the GPUs on for another few months.

1:56 But, this lifeline has, in turn,

1:58 presented another potential issue that would almost sound absurd if it wasn't

2:02 for a growing pool of people raising the alarm bells over it.

2:05 OpenAI, alongside some other big names,

2:07 have now become so valuable in private markets

2:10 that if they do all go public this year,

2:12 their collective waiting could legitimately break the stock market.

2:16 OpenAI, as you rightly pointed out,

2:18 had talked about almost 1.4 trillion dollars in investments.

2:23 Google, Amazon, Microsoft, the video,

2:25 they're all kind of backing everyone right now to lift up the wider AI market.

2:29 OpenAI declaring code red as the AI race intensifies,

2:34 and Google threatens to unseat [music] the industry's early leader.

2:37 These US big tech corporations are fearmongering about China.

2:41 That raises a lot of questions about how OpenAI employees feel about this.

2:45 Workers using AI are more productive.

2:48 They're also more burnt out than ever before.

2:51 [music] Brad, if you want to sell your shares, I'll find you a buyer.

2:55 Now, despite how it may appear from the outside,

2:57 the business leaders and investment managers pouring

3:00 hundreds of billions of dollars into OpenAI

3:02 are fully aware of the questions and criticisms people have about the business.

3:06 It's not like every single person on the internet has seen something

3:09 that the CEO of Microsoft or Nvidia somehow didn't get the memo for.

3:12 They're all capable of doing the basic arithmetic that concludes

3:16 that a lot of these numbers don't really make a lot of sense,

3:18 but they are still betting on it anyway.

3:21 By the business fundamentals alone, yeah, it doesn't look great.

3:24 But, to these mega investors, the case for OpenAI can continue to be

3:28 justified so long as four things remain true.

3:31 Now, unfortunately for them, out of those four variables,

3:34 there are only three that the company can really control.

3:37 It has to maintain extremely strong user growth,

3:39 preferably amongst users that are actually willing to pay money for the service,

3:42 it has to demonstrate that those customers will actually

3:45 stick around to be profited off of, and it

3:47 has to remain a viable contender for the long

3:49 shot bet of developing genuinely world-changing superintelligence.

3:53 If the company can show investors that a large group of people

3:55 are willing to pay a recurring subscription fee for their services,

3:58 it is a good indication that there may be a path to profitability in the future,

4:02 even if they are losing a lot of money

4:03 at the moment for every customer they serve,

4:05 even those in the top subscription tiers.

4:08 If they can fully capture that market,

4:10 they can basically try to do what every tech

4:12 company ever has done through a gradual process of enshittification.

4:16 The game plan is simple.

4:18 They start out with a really useful and cost-effective service.

4:21 Get people used to using it, drive out all competition,

4:24 and once they have done that, they can raise prices,

4:26 cut costs, and integrate other revenue streams like ads

4:29 to profit off the user base they have cornered.

4:31 For technology in particular,

4:32 this is a very effective strategy to build companies because

4:35 the marginal cost of every additional user is effectively zero.

4:39 So, the benefits of being the sole player in the market is enormous.

4:42 A company like OpenAI also has an advantage here because over time

4:46 the cost to deliver the same service

4:48 should come down naturally as technology improves.

4:51 It may not feel like it at the moment,

4:53 especially if you have tried to build a computer recently,

4:56 but the cost of the hardware and overhead

4:58 needed to run AI is coming down extremely quickly.

5:01 Tokens are the basic building blocks of AI-generated content,

5:04 so [music] it's easiest to track them as a point of comparison.

5:07 According to data from the AI Index Report prepared by Epic AI,

5:10 when ChatGPT first launched in late 2022,

5:13 the cost to produce a million tokens was around $11.

5:16 In early 2025, that cost had come all the way down

5:19 to just 9 cents with some models from competitors running even cheaper.

5:24 Now, that is the most recent data that has been reliably reported

5:27 on, but the trend is that since the very early generative models,

5:30 the cost per token has been halving 2.7 times over every single year.

5:34 The only reason why spending has increased so much overall is

5:37 because OpenAI and its competitors are spending a lot on training larger,

5:40 more capable models, and we are just using a lot more of these tokens overall.

5:45 The data is hard to collect, but according to the tracking platform OpenRouter,

5:49 the number of tokens consumed rose by over 38

5:52 times over a 12-month period ending late last year.

5:55 This is all to say that even if

5:57 AI remains nothing more than a little handy tool,

5:59 the demand for those tools is rising,

6:01 and the cost to deliver them is falling just as quickly.

6:04 Additionally, the CEOs are betting that once

6:06 a powerful enough model is developed,

6:08 they won't need to keep investing the enormous amounts

6:10 of money they have been in the early development stage.

6:13 That means if one of these businesses can corner the market,

6:16 they could make a lot of money with something as simple as a subscription model,

6:19 or yes, even advertising.

6:21 If an OpenAI subscription becomes as ubiquitous as a Netflix subscription,

6:25 then that alone can justify the value of these investments.

6:27 Because Netflix, by comparison, does not have the luxury of movies

6:30 becoming 27 times cheaper to produce every year.

6:34 This is not even considering the revenue

6:35 that could come from commercial or government clients.

6:37 But well, yeah, we will get to that soon.

6:40 This is also ignoring the long-shot bet that OpenAI is, or at least was,

6:44 the front runner for developing some kind of super

6:46 intelligence with capabilities worth tens of trillions of dollars.

6:50 A lot of real experts on this topic have pointed

6:52 out that this kind of scenario is extremely unlikely to happen.

6:55 But even if there is only a one in 1,000 chance,

6:58 if that investment could pay out 10,000 times over, it's still a good bet.

7:02 Especially if the backup plan is merely becoming

7:05 one of the most valuable SaaS companies in existence.

7:08 It's a great story that almost makes it sound like these big

7:10 tech CEOs may not be intentionally lighting their money on fire after all.

7:14 But before you personally get your checkbook out,

7:17 it's time to learn how many words to find

7:18 out why this story may be starting to fall apart.

7:21 It's not just the companies that you hand

7:23 your data to willingly that you need to worry about.

7:25 Hundreds of data brokers are collecting

7:27 and selling your personal information right now,

7:29 and most people have no idea it's happening.

7:32 That's why this video is sponsored by Incogni.

7:34 Your email addresses, phone numbers, home address,

7:37 and more are probably sitting on data broker

7:39 websites ready for scammers and identity thieves to buy.

7:42 And it's your fault.

7:43 It's mostly because the companies you trusted

7:45 with your data weren't careful enough with it.

7:47 Incogni automatically contacts these data brokers

7:49 on your behalf and demands they remove your information.

7:53 They handle the back and forth, the follow-ups,

7:55 and the legal requests so you don't have to.

7:58 They can even handle custom removals for specific

8:00 instances of your info being leaked online.

8:02 Just Google yourself and you might be

8:04 surprised to find that new sites, fitness apps,

8:06 or random websites you've never heard

8:08 of are holding and sharing your personal details.

8:10 And Incogni will go after those individually, too.

8:13 The way I see it, they can't steal

8:15 your identity if they have no idea who you are.

8:18 Go to incogni.com/howmanyworks or use code howmanyworks at the link below

8:22 to get 60% off an annual plan and take your personal data back.

8:27 Okay.

8:28 So, behind the scenes, some of the optimistic outcomes for these variables

8:31 has not been panning out too well for OpenAI.

8:34 And of course, a lot of those challenges were brought to the surface last week.

8:37 If you are not up to speed, the brief summary is that Anthropic,

8:41 the company behind Claude and OpenAI's chief rival,

8:43 pushed back against the Department of War over using

8:45 their technology for fully autonomous armed drones and mass surveillance.

8:49 Now, we are already a long way beyond Asimov's three laws of robotics here,

8:53 but many people were still willing to give credit to Anthropic

8:55 for doing what you might reasonably consider to be the bare minimum,

8:59 especially when they were put under significant pressure from the government.

9:02 Now, all of that public good was very quickly turned into animosity when just

9:06 a few hours later OpenAI turned around

9:08 and signed a similar deal with the Pentagon,

9:11 effectively announcing that they were willing to forego these extremely basic AI

9:14 safety guardrails if it meant that they

9:16 could secure a lucrative government contract.

9:18 Now, I respect your time and I know most of you are probably

9:21 already up to speed on this, but if you do want more details,

9:24 I will leave a link to some articles in the description,

9:26 as well as some great detailed commentary done by Mr.

9:29 Glizzy Hands himself.

9:30 Now, the whole situation is very interesting and frankly

9:33 terrifying from the ethics side of the AI debate,

9:36 but it has also been quite revealing from the business side as well.

9:39 Following this episode, a lot of people are actively

9:42 boycotting OpenAI's products and considering alternatives,

9:45 with the biggest winner so far being Anthropic itself.

9:47 This undermines the narrative of continual user growth,

9:50 and it highlights that people are not really

9:52 that attached to any particular brand of AI.

9:55 For the past 2 years, OpenAI in particular has been able to command

9:58 an investment premium in part because it was the first thing that people thought

10:01 of when it came to large language models.

10:04 It was the Google Coca-Cola, Band-Aid, Clorox, and Velcro of AI tools.

10:08 To a lot of average people, ChatGPT was just the way to use AI.

10:13 It was also assumed that once people got used to using a particular model,

10:16 it would be very hard to get them to switch,

10:18 which in turn would make it easier to generate

10:21 profit through higher fees and a cheaper service.

10:23 This whole Killer Robots episode may not single-handedly destroy OpenAI,

10:27 but it has been a very visible demonstration that people

10:30 are happy to switch if they see a reason to.

10:33 Competitors have capitalized on this by releasing tools

10:35 to make this switch between models as seamless as possible,

10:38 even for users who had a lot of custom trained data built up with OpenAI.

10:42 This has hit two of the three stories that OpenAI

10:45 needed to keep going in order to maintain investor interest,

10:48 and it also hit this last one as well.

10:50 The race to secure talent in the AI space has become somewhat ludicrous,

10:54 with pay packets into the tens of millions of dollars a year.

10:58 But, companies are hoping that these salaries will be worth

11:00 it if it improves their chances of making a general intelligence.

11:04 The blowback over these ethical decisions will make

11:06 it harder for OpenAI to attract these high-end researchers,

11:09 and it will also make it more likely that the ones

11:11 they currently have will look for opportunities with competitors.

11:14 Now, of course, some of these people may not really care about

11:17 what their AI is doing as long as the paychecks come in.

11:20 But, with multi-million dollar signing bonuses being thrown around,

11:22 at the very least these ethical concerns are a great excuse to jump ship.

11:27 Big investors care about these talent movements

11:29 because the The these people move around,

11:31 the more the institutional knowledge gets spread around.

11:34 If a highly skilled developer spends two

11:35 years at OpenAI and then moves to Anthropic,

11:38 they are going to bring with them some

11:39 level of know-how about how to develop similar capabilities.

11:42 Now, obviously, these people will be under very strict non-disclosure contracts,

11:46 but even without copying anything directly,

11:48 these programs are ultimately the product of the people who

11:50 designed them and a whole lot of stolen intellectual property.

11:54 But, we don't talk about that part out loud.

11:56 These knowledge leaks make it more likely that AI

11:59 will just become that any somewhat competent business

12:02 can create rather than a centralized product that people

12:04 need to pay for the privilege of using.

12:07 Now, this isn't exactly unprecedented in the world of technology, either.

12:10 For all of the emphasis that is put on being a pioneer in a new field,

12:14 it rarely, if ever, guarantees business success.

12:17 Nobody uses AltaVista, BlackBerry, MySpace, or Skype anymore,

12:20 despite them all at one time being the cutting edge of their respective fields.

12:24 Now, for a company like Google, Meta, Microsoft, Amazon,

12:27 or even xAI, this isn't as big a deal.

12:30 If AI does become a universally accessible commodity,

12:33 then all of the money they have spent

12:34 building their own capabilities may have been a waste.

12:37 But, at least the cost-cutting made possible

12:39 by this technology could still help their core business functions.

12:42 This is, in part, why companies like Google and Amazon are

12:45 developing their own AI products while

12:47 simultaneously investing in their own competition.

12:50 However, for a company like OpenAI,

12:52 if they can't monetize the technology itself, they are in big trouble,

12:56 especially since they have made some of the most significant

12:58 upfront investments to develop this whole industry in the first place.

13:02 An analogy that has been thrown around is that they are like a cheetah

13:04 that has used a lot of resources to hunt down this new technology.

13:07 And now that they have, the bigger,

13:09 slower predators can come along and steal the rewards

13:11 without having to use as much energy of their own.

13:14 Now, of course, the other major tech companies have

13:16 invested a lot of money into pursuing this technology,

13:19 but it was largely money that they already had,

13:21 and they have other profit centers to keep

13:23 this going even if investors get cold feet,

13:25 which brings up that one big variable they can't control.

13:29 The big investors can keep justifying the money they spend on Open

13:31 AI so long as their investors see the potential for the technology,

13:35 or at the very least see the opportunity for the hype to generate some kind

13:39 of a return on their investment as they sell

13:40 their stake off to the next person in line.

13:42 Investor enthusiasm around AI is already waning and market

13:46 conditions alongside higher interest rates aren't helping either.

13:49 But the ultimate light at the end

13:50 of the tunnel for a lot of these early investors,

13:53 as well as the employees paid in company stock,

13:55 is to take the company public so that they

13:57 can cash out their positions to the general public.

13:59 Now, so far the company has only teased the idea

14:02 of an IPO with no formal plans actively submitted.

14:05 But it is still worth considering how such a move

14:07 might play out because whether you like it or not,

14:09 it's probably going to affect you.

14:11 In any scenario, an initial public offering

14:13 brings a lot of additional scrutiny onto

14:15 a company as its financials and operations are

14:17 thoroughly picked apart in the due diligence process.

14:20 This has been the undoing of several hyped companies in the past,

14:23 most famously WeWork, a business built on bold promises that didn't quite

14:26 live up to the optimism of the eccentric founder,

14:29 also backed by major investments from SoftBank.

14:32 Now, honestly, this is probably a bit unfair.

14:34 Open AI is at the very least a real technology company,

14:37 not an office subletter that pretends to be a technology company.

14:40 But it will still present a hurdle for a slightly different reason.

14:43 Going public means that the company leadership will be subject

14:46 to a lot more regulation about the stories they are allowed to tell.

14:49 Altman in particular has gained a bit of a reputation

14:52 for making big promises about the financial potentials of his businesses,

14:55 including those that he ran before Open AI.

14:58 This kind of messaging will need to be picked a lot more

15:00 carefully in the run-up to a listing and after the company is public.

15:04 Now, I know what you might be thinking.

15:06 A lot of public company CEOs have made a mockery of these regulations,

15:09 and enforcement by bodies like the SEC has effectively amounted

15:13 to a slap on the wrist with a wet noodle.

15:15 But for such an important business,

15:17 it's not impossible that enforcement action could be

15:20 used against them punitively if they don't play along.

15:23 I mean, it literally happened just last week.

15:25 So, yeah, OpenAI has made promises that financially cannot afford

15:29 to keep for technology that is no longer special being offered

15:32 to users that are moving to competitors and talent that is

15:34 being spread around to higher bidders that come with less reputational baggage.

15:38 As these problems have caught up with the company,

15:40 it has tried things that are looking increasingly desperate

15:43 in order to keep the numbers moving in the right direction.

15:45 Goonbots, integrated advertising, and questionable military contracts may

15:49 have introduced additional revenue streams,

15:51 but they made all of these fundamental challenges even worse.

15:54 It sounds bad, but I guess we should talk

15:56 about how this company might uh break the stock market.

16:00 So, despite all of these problems,

16:02 OpenAI could still easily be the largest IPO in history.

16:05 After the most recent a hundred and ten billion dollar investment,

16:08 the company claimed a pre-money valuation

16:10 of seven hundred and thirty billion dollars,

16:12 which already makes it one of the most valuable companies

16:14 in the world slotting right in between JP Morgan Chase and Exxon.

16:18 Now, not all shares would be made instantly

16:20 available on the market on the first day,

16:22 but even assuming an extremely conservative

16:24 five percent of the company's shares,

16:25 it would still be close to forty billion dollars in float,

16:28 topping even Saudi Aramco's IPO,

16:30 which was a state oil company and not really a fair comparison.

16:34 [music] As a closer comparison, when Facebook went public in 2012,

16:36 it put up fifteen percent of its

16:38 stock to raise an inflation-adjusted twenty-two billion dollars.

16:41 A similar offering from OpenAI would be five times that.

16:45 Now, the market is obviously a lot bigger today than it was fourteen years ago,

16:48 but the majority of net share purchases are

16:51 now done by companies themselves buying their own stock.

16:54 An IPO this big would represent a significant

16:56 share of the actual new investment into the market,

16:59 theoretically meaning that there would be

17:00 less investor dollars for everything else.

17:02 Now, that might be a problem for the market, not OpenAI, but at the same time,

17:06 there are other big IPOs rumored from SpaceX and potentially even Anthropic.

17:11 Collectively, these companies could raise more money this year

17:13 than the last decade of IPO activity combined.

17:16 That cash has to come from somewhere,

17:19 and the fear is that the market just won't be able

17:20 to bear this much supply in such a short period of time.

17:24 According to the Z1 accounts of financial activity,

17:26 the $450 billion that just these companies could look to raise would be more

17:30 than the net buying activity seen in US markets in a lot of years,

17:34 and that's ignoring every single other company

17:36 that people might want to invest in.

17:38 If OpenAI can't raise this cash, it will put more question marks

17:41 over its future and proposed spending commitments.

17:44 The sheer size of these potential new

17:45 public companies is also raising another concern,

17:48 which is that large indexes that track the stock

17:50 market will have to readjust to include them.

17:52 Index funds from companies like Vanguard, BlackRock,

17:55 and State Street automatically adjust their holdings to weight their indexes

17:58 proportionally to the value of the companies within that index.

18:01 The most valuable companies in the world today all started

18:04 out as relatively small-cap stocks that just kept on growing,

18:07 giving these indexes time to weight them more heavily as they gradually grew.

18:11 If these three IPOs all go ahead at their proposed values,

18:14 they would instantly jump close to the top of these indexes,

18:17 which could force these large asset managers to sell down a lot

18:20 of other companies in order to correctly

18:21 weight their indexes to reflect the market.

18:24 Now, not to get too nerdy,

18:25 but if only 10% of these companies are actively sold on the market,

18:28 then most indexes are float-adjusted,

18:30 meaning [music] that only 10% of the company

18:32 would be counted towards overall holdings.

18:34 But even this could present a major shake-up in a short period of time.

18:38 My good friend Ben Felix actually spoke directly to Vanguard's global

18:41 head of investment implementation in a podcast he did last year.

18:44 So, if you want to get really nerdy,

18:46 they go into a lot of the technical details about how this could play out,

18:49 and I will leave a link to that in the description below.

18:52 Now, nobody really knows exactly how these challenges will play out,

18:56 including the company themselves.

18:58 But unfortunately, the trend is towards a majority

19:00 of this risk being dumped onto public investors, or just the public in general.

19:04 If these companies fail, they could bring down the market with them.

19:07 But if they succeed, it will likely be because they were used

19:10 to displace a large share of the workforce.

19:12 If the optimistic outcome is a majority of human work becoming redundant,

19:16 go and watch this video next to find out why these very

19:19 same tech bros want you to have as many children as possible.

19:22 And don't forget to like and subscribe to keep on learning how money works.