0:02 Open AAI has not been having a great
0:04 year. As it completes its metamorphosis
0:06 from a nonprofit organization operating
0:08 for the good of humanity to a
0:10 shareholder focused company gearing up
0:12 for one of the biggest IPOs in history,
0:15 it has come up against some uh teething
0:17 problems. It is facing legal challenges
0:18 over the fair use of intellectual
0:20 property, legal challenges over its
0:22 governance, legal challenges over the
0:24 safe use of its tools, talent churn
0:26 amongst its top developers, and now
0:28 significant reputational damage amongst
0:30 effectively all of its stakeholders. All
0:32 of this wasn't helped by its willingness
0:34 to embrace what many see as war
0:36 profitering, further straying away from
0:37 their original stated goal in order to
0:40 chase revenue wherever it can find it.
0:42 because well I mean fully autonomous AI
0:44 killbots and mass surveillance systems
0:45 don't really sound like they are
0:47 improving the future of humanity. In
0:49 response to this over just the last
0:51 weeks some of even the most staunch AI
0:53 optimists have spoken out against the
0:55 decisions this company has been making.
0:58 But anyway, all of this is on top of the
0:59 biggest and most immediate problem of
1:01 all, which is that OpenAI is simply
1:03 burning billions of dollars every month
1:05 on top of trillions of dollars in future
1:07 spending commitments with no clear
1:09 answer for how it's going to get that
1:12 money. Now, maybe this would all be okay
1:14 if it was still the undisputed leader in
1:16 this world changing technology. But in
1:18 the last 3 years, several other models
1:20 from better funded companies have come
1:22 along and either matched or exceeded
1:24 OpenAI's capabilities. It's now unclear
1:26 if the early pioneers of this technology
1:28 have spent hundreds of billions of
1:29 dollars developing systems that can now
1:31 be effectively replicated by random
1:33 Swedish men in their bedroom. It's also
1:35 becoming less clear exactly how
1:37 worldchanging this technology really is
1:39 going to be in the first place, putting
1:40 more question marks over the company's
1:42 future. For now, its investors are
1:45 doubling down. Just last week, Amazon,
1:46 who is theoretically a competitor,
1:48 Nvidia, a supplier, and the ever
1:50 reliable Soft Bank, announced a record
1:53 investment of $110 billion in additional
1:55 funding, which will keep the GPUs on for
1:57 another few months. But this lifeline
2:00 has in turn presented another potential
2:02 issue that would almost sound absurd if
2:03 it wasn't for a growing pool of people
2:06 raising the alarm bells over it. OpenAI
2:08 alongside some other big names have now
2:10 become so valuable in private markets
2:12 that if they do all go public this year,
2:14 their collective waiting could
2:16 legitimately break the stock market.
2:18 >> Open AAI, as you rightly pointed out,
2:22 had talked about almost $1.4 trillion in investments.
2:23 investments.
2:25 >> Google, Amazon, Microsoft, Nvidia,
2:27 they're all kind of backing everyone
2:29 right now to lift up the wider AI market.
2:29 market.
2:32 >> Open AI declaring code red as the AI
2:35 race intensifies. and Google threatens
2:37 to unseat the industry's early leader.
2:39 >> These US big tech corporations are
2:41 fear-mongering about China.
2:43 >> That raises a lot of questions about how
2:45 AI employees feel about this.
2:48 >> Workers using AI are more productive.
2:51 They're also more burned out than ever before.
2:51 before.
2:53 >> Brad, if you want to sell your shares,
2:55 I'll find you a buyer.
2:57 Now, despite how it may appear from the
2:59 outside, the business leaders and
3:00 investment managers pouring hundreds of
3:02 billions of dollars into Open AI are
3:04 fully aware of the questions and
3:05 criticisms people have about the
3:07 business. It's not like every single
3:08 person on the internet has seen
3:10 something that the CEO of Microsoft or
3:12 Nvidia somehow didn't get the memo for.
3:14 They are all capable of doing the basic
3:16 arithmetic that concludes that a lot of
3:18 these numbers don't really make a lot of
3:20 sense, but they are still betting on it
3:22 anyway. By the business fundamentals
3:25 alone, yeah, it doesn't look great. But
3:26 to these mega investors, the case for
3:29 OpenAI can continue to be justified so
3:32 long as four things remain true. Now,
3:33 unfortunately for them, out of those
3:35 four variables, there are only three
3:37 that the company can really control. It
3:39 has to maintain extremely strong user
3:40 growth, preferably amongst users that
3:42 are actually willing to pay money for
3:44 the service. It has to demonstrate that
3:45 those customers will actually stick
3:47 around to be profited off of. And it has
3:49 to remain a viable contender for the
3:51 longot bed of developing genuinely
3:53 worldchanging super intelligence. If the
3:55 company can show investors that a large
3:56 group of people are willing to pay a
3:58 recurring subscription fee for their
3:59 services, it is a good indication that
4:01 there may be a path to profitability in
4:03 the future. Even if they are losing a
4:04 lot of money at the moment for every
4:06 customer they serve, even those in the
4:08 top subscription tiers. If they can
4:10 fully capture that market, they can
4:12 basically try to do what every tech
4:14 company ever has done through a gradual
4:17 process of inshitification. The game
4:19 plan is simple. They start out with a
4:20 really useful and cost-effective
4:22 service. get people used to using it,
4:24 drive out all competition, and once they
4:26 have done that, they can raise prices,
4:28 cut costs, and integrate other revenue
4:30 streams like ads to profit off the user
4:32 base they have cornered. For technology
4:34 in particular, this is a very effective
4:36 strategy to build companies because the
4:38 marginal cost of every additional user
4:40 is effectively zero. So, the benefit of
4:41 being the sole player in the market is
4:44 enormous. A company like OpenAI also has
4:47 an advantage here because over time the
4:48 cost to deliver the same service should
4:50 come down naturally as technology
4:53 improves. It may not feel like it at the
4:54 moment, especially if you have tried to
4:56 build a computer recently, but the cost
4:58 of the hardware and overhead needed to
5:01 run AI is coming down extremely quickly.
5:03 Tokens are the basic building blocks of
5:05 AI generated content, so it's easiest to
5:07 track them as a point of comparison.
5:09 According to data from the AI index
5:11 report prepared by Epic AI when Chad GPT
5:14 first launched in late 2022, the cost to
5:16 produce a million tokens was around $11.
5:19 In early 2025, that cost had come all
5:21 the way down to just 9 with some models
5:24 from competitors running even cheaper.
5:25 Now, that is the most recent data that
5:28 has been reliably reported on. But the
5:29 trend is that since very early
5:31 generative models, the cost per token
5:33 has been having 2.7 times over every
5:35 single year. The only reason why
5:37 spending has increased so much overall
5:39 is because OpenAI and its competitors
5:41 are spending a lot on training larger,
5:43 more capable models. And we are just
5:44 using a lot more of these tokens
5:47 overall. The data is hard to collect,
5:48 but according to the tracking platform
5:50 Open Router, the number of tokens
5:52 consumed rose by over 38 times over a
5:55 12-month period ending late last year.
5:57 This is all to say that even if AI
5:59 remains nothing more than a little handy
6:00 tool, the demand for those tools is
6:02 rising and the cost to deliver them is
6:04 falling just as quickly. Additionally,
6:06 the CEOs are betting that once a
6:08 powerful enough model is developed, they
6:09 won't need to keep investing the
6:11 enormous amounts of money they have been
6:13 in the early development stage. That
6:15 means if one of these businesses can
6:16 corner the market, they could make a lot
6:18 of money with something as simple as a
6:20 subscription model or yes, even
6:22 advertising. If an OpenAI subscription
6:24 becomes as ubiquitous as a Netflix
6:25 subscription, then that alone could
6:27 justify the value of these investments
6:29 because Netflix by comparison does not
6:31 have the luxury of movies becoming 27
6:33 times cheaper to produce every year.
6:35 This is not even considering the revenue
6:36 that could come from commercial or
6:39 government clients. But well, yeah, we
6:41 will get to that soon. This is also
6:43 ignoring the longshot bet that OpenAI
6:45 is, or at least was the front runner for
6:46 developing some kind of super
6:48 intelligence with capabilities worth
6:50 tens of trillions of dollars. A lot of
6:52 real experts on this topic have pointed
6:53 out that this kind of scenario is
6:55 extremely unlikely to happen. But even
6:58 if there is only a 1 in 1,000 chance if
7:00 that investment could pay out 10,000
7:02 times over, it's still a good bet.
7:04 Especially if the backup plan is merely
7:06 becoming one of the most valuable SAS
7:08 companies in existence. It's a great
7:10 story that almost makes it sound like
7:12 these big tech CEOs may not be
7:13 intentionally lighting their money on
7:15 fire after all. But before you
7:17 personally get your checkbook out, it's
7:18 time to learn how many works to find out
7:20 why this story may be starting to fall
7:22 apart. It's not just the companies that
7:24 you hand your data to willingly that you
7:26 need to worry about. Hundreds of data
7:28 brokers are collecting and selling your
7:30 personal information right now, and most
7:32 people have no idea it's happening.
7:33 That's why this video is sponsored by
7:36 Incogn. Your email addresses, phone
7:37 numbers, home address, and more are
7:39 probably sitting on data broker websites
7:41 ready for scammers and identity thieves
7:43 to buy. And it's not your fault. It's
7:45 mostly because the companies you trusted
7:46 with your data weren't careful enough
7:48 with it. And Cognney automatically
7:50 contacts these data brokers on your
7:52 behalf and demands they remove your
7:54 information. They handle the back and
7:56 forth, the follow-ups, and the legal
7:58 requests so you don't have to. They can
8:00 even handle custom removals for specific
8:02 instances of your info being leaked
8:04 online. Just Google yourself, and you
8:05 might be surprised to find that news
8:07 sites, fitness apps, or random websites
8:09 you've never heard of are holding and
8:11 sharing your personal details. Incogn
8:13 will go after those individually, too.
8:15 The way I see it, they can't steal your
8:17 identity if they have no idea who you
8:20 are. Go to incogn.com/howmoney
8:22 works or use code how money works at the
8:25 link below to get 60% off an annual plan
8:28 and take your personal data back. Okay,
8:29 so behind the scenes, some of the
8:31 optimistic outcomes for these variables
8:33 has not been panning out too well for
8:35 OpenAI. And of course, a lot of those
8:37 challenges were brought to the surface
8:39 last week. If you are not up to speed,
8:41 the brief summary is that Anthropic, the
8:43 company behind Claude and OpenAI's chief
8:44 rival, pushed back against the
8:46 Department of War over using their
8:47 technology for fully autonomous armed
8:49 drones and mass surveillance. Now, we
8:52 are already a long way beyond Asimov's
8:53 three laws of robotics here. But many
8:55 people were still willing to give credit
8:57 to Anthropic for doing what you might
8:59 reasonably consider to be the bare
9:00 minimum, especially when they were put
9:02 under significant pressure from the
9:04 government. Now, all of that public good
9:06 was very quickly turned into animosity
9:08 when just a few hours later, OpenAI
9:10 turned around and signed a similar deal
9:11 with the Pentagon, effectively
9:12 announcing that they were willing to
9:15 forgo these extremely basic AI safety
9:16 guardrails if it meant that they could
9:18 secure a lucrative government contract.
9:20 Now, I respect your time, and I know
9:22 most of you are probably already up to
9:23 speed on this, but if you do want more
9:25 details, I will leave a link to some
9:27 articles in the description, as well as
9:28 some great detailed commentary done by
9:31 Mr. Gizzy Hands himself. Now, the whole
9:33 situation is very interesting and
9:35 frankly terrifying from the ethics side
9:37 of the AI debate, but it has also been
9:38 quite revealing from the business side
9:41 as well. Following this episode, a lot
9:42 of people are actively boycotting
9:44 OpenAI's products and considering
9:46 alternatives with the biggest winner so
9:48 far being Anthropic itself. This
9:49 undermines the narrative of continual
9:51 user growth and it highlights that
9:53 people are not really that attached to
9:55 any particular brand of AI. For the past
9:57 two years, OpenAI in particular has been
9:59 able to command an investment premium in
10:01 part because it was the first thing that
10:03 people thought of when it came to large
10:05 language models. It was the Google,
10:07 Coca-Cola, Band-Aid, Clorox, and Velcro
10:10 of AI tools. To a lot of average people,
10:14 chat GPT was just the way to use AI. It
10:15 was also assumed that once people got
10:17 used to using a particular model, it
10:18 would be very hard to get them to
10:20 switch, which in turn would make it
10:22 easier to generate profit through higher
10:24 fees and a cheaper service. This whole
10:25 killer robots episode may not
10:28 single-handedly destroy open AI, but it
10:29 has been a very visible demonstration
10:32 that people are happy to switch if they
10:34 see a reason to. Competitors have
10:35 capitalized on this by releasing tools
10:37 to make this switch between models as
10:39 seamless as possible, even for users who
10:41 had a lot of custom trained data built
10:44 up with OpenAI. This has hit two of the
10:46 three stories that OpenAI needed to keep
10:47 going in order to maintain investor
10:50 interest. And it also hit this last one
10:52 as well. The race to secure talent in
10:54 the AI space has become somewhat
10:56 ludicrous with pay packets into the tens
10:58 of millions of dollars a year. But
11:00 companies are hoping that these salaries
11:01 will be worth it if it improves their
11:03 chances of making a general
11:05 intelligence. The blowback over these
11:06 ethical decisions will make it harder
11:08 for OpenAI to attract these high-end
11:10 researchers and it will also make it
11:11 more likely that the ones they currently
11:13 have will look for opportunities with
11:15 competitors. Now, of course, some of
11:17 these people may not really care about
11:19 what their AI is doing as long as the
11:20 paychecks come in, but with
11:22 multi-million dollar signing bonuses
11:24 being thrown around, at the very least,
11:26 these ethical concerns are a great
11:28 excuse to jump ship. Big investors care
11:30 about these talent movements because the
11:31 more these people move around, the more
11:33 the institutional knowledge gets spread
11:35 around. If a highly skilled developer
11:37 spends 2 years at OpenAI and then moves
11:38 to Enthropic, they are going to bring
11:40 with them some level of knowhow about
11:43 how to develop similar capabilities. Now
11:44 obviously these people will be under
11:46 very strict non-disclosure contracts but
11:48 even without copying anything directly
11:49 these programs are ultimately the
11:51 product of the people who designed them
11:53 and a whole lot of stolen intellectual
11:55 property but we don't talk about that
11:57 part out loud. These knowledge leaks
11:59 make it more likely that AI will just
12:01 become a commodity that any somewhat
12:03 competent business can create rather
12:05 than a centralized product that people
12:07 need to pay for the privilege of using.
12:09 Now, this isn't exactly unprecedented in
12:11 the world of technology either. For all
12:12 of the emphasis that is put on being a
12:15 pioneer in a new field, it rarely if
12:17 ever guarantees business success. Nobody
12:20 uses Alta Vista, Blackberry, MySpace, or
12:22 Skype anymore, despite them all at one
12:23 time being the cutting edge of their
12:25 respective fields. Now, for a company
12:27 like Google, Meta, Microsoft, Amazon, or
12:30 even XAI, this isn't as big a deal. If
12:33 AI does become a universally accessible
12:34 commodity, then all of the money they
12:35 have spent building their own
12:37 capabilities may have been a waste. But
12:39 at least the cost cutting made possible
12:40 by this technology could still help
12:43 their core business functions. This is
12:45 in part why companies like Google and
12:46 Amazon are developing their own AI
12:48 products while simultaneously investing
12:51 in their own competition. However, for a
12:53 company like OpenAI, if they can't
12:55 monetize the technology itself, they are
12:57 in big trouble, especially since they
12:58 have made some of the most significant
13:00 upfront investments to develop this
13:02 whole industry in the first place. An
13:03 analogy that has been thrown around is
13:05 that they are like a cheetah that has
13:06 used a lot of resources to hunt down
13:08 this new technology. And now that they
13:10 have, the bigger, slower predators can
13:12 come along and steal the rewards without
13:13 having to use as much energy of their
13:16 own. Now, of course, the other major
13:17 tech companies have invested a lot of
13:19 money into pursuing this technology, but
13:21 it was largely money that they already
13:22 had, and they have other profit centers
13:24 to keep this going, even if investors
13:27 get cold feet, which brings up that one
13:29 big variable they can't control. The big
13:31 investors can keep justifying the money
13:33 they spend on Open AI so long as their
13:34 investors see the potential for the
13:37 technology, or at the very least see the
13:38 opportunity for the hype to generate
13:39 some kind of a return on their
13:41 investment as they sell their stake off
13:43 to the next person in line. Investor
13:45 enthusiasm around AI is already waning
13:47 and market conditions alongside higher
13:49 interest rates aren't helping either.
13:51 But the ultimate light at the end of the
13:52 tunnel for a lot of these early
13:54 investors as well as the employees paid
13:56 in company stock is to take the company
13:58 public so that they can cash out their
14:00 positions to the general public. Now, so
14:02 far the company has only teased the idea
14:04 of an IPO with no formal plans actively
14:06 submitted. But it is still worth
14:07 considering how such a move might play
14:09 out because whether you like it or not,
14:12 it's probably going to affect you. In
14:13 any scenario, an initial public offering
14:15 brings a lot of additional scrutiny onto
14:16 a company as its financials and
14:18 operations are thoroughly picked apart
14:20 in the due diligence process. This has
14:22 been the undoing of several hyped
14:23 companies in the past, most famously
14:25 WeWork, a business built on bold
14:27 promises that didn't quite live up to
14:29 the optimism of the eccentric founder,
14:31 also backed by major investments from
14:33 SoftBank. Now, honestly, this is
14:35 probably a bit unfair. Open AAI is at
14:37 the very least a real technology
14:39 company, not an office subletter that
14:41 pretends to be a technology company, but
14:42 it will still present a hurdle for a
14:44 slightly different reason. Going public
14:46 means that the company leadership will
14:47 be subject to a lot more regulation
14:49 about the stories they are allowed to
14:51 tell. Alman, in particular, has gained a
14:53 bit of a reputation for making big
14:54 promises about the financial potentials
14:56 of his businesses, including those that
14:59 he ran before OpenAI. This kind of
15:00 messaging will need to be picked a lot
15:02 more carefully in the run-up to a
15:04 listing and after the company is public.
15:06 Now, I know what you might be thinking.
15:08 A lot of public company CEOs have made a
15:10 mockery of these regulations, and
15:12 enforcement by bodies like the SEC has
15:13 effectively amounted to a slap on the
15:16 wrist with a wet noodle. But for such an
15:18 important business, it's not impossible
15:20 that enforcement action could be used
15:22 against them punitively if they don't
15:24 play along. I mean, it literally
15:27 happened just last week. So yeah, OpenAI
15:28 has made promises that financially
15:30 cannot afford to keep for technology
15:32 that is no longer special, being offered
15:34 to users that are moving to competitors
15:35 and talent that is being spread around
15:37 to higher biders that come with less
15:39 reputational baggage. As these problems
15:41 have caught up with the company, it has
15:42 tried things that are looking
15:44 increasingly desperate in order to keep
15:45 the numbers moving in the right
15:47 direction. Goonbots, integrated
15:49 advertising, and questionable military
15:50 contracts may have introduced additional
15:52 revenue streams, but they made all of
15:54 these fundamental challenges even worse.
15:56 It sounds bad, but I guess we should
15:59 talk about how this company might uh
16:01 break the stock market. So, despite all
16:03 of these problems, OpenAI could still
16:05 easily be the largest IPO in history.
16:07 After the most recent $110 billion
16:09 investment, the company claimed a pre-
16:12 money valuation of $730 billion, which
16:13 already makes it one of the most
16:15 valuable companies in the world,
16:17 slotting right in between JP Morgan,
16:19 Chase, and Exxon. Now, not all shares
16:21 would be made instantly available on the
16:22 market on the first day, but even
16:24 assuming an extremely conservative 5% of
16:26 the company's shares, it would still be
16:28 close to $40 billion in float, topping
16:30 even Sadia Ramco's IPO, which was a
16:32 state oil company and not really a fair
16:35 comparison. As a closer comparison, when
16:37 Facebook went public in 2012, it put up
16:38 15% of its stock to raise an
16:41 inflationadjusted $22 billion. A similar
16:44 offering from OpenAI would be five times
16:46 that. Now, the market is obviously a lot
16:49 bigger today than it was 14 years ago,
16:51 but the majority of net share purchases
16:53 are now done by companies themselves
16:55 buying their own stock. An IPO this big
16:57 would represent a significant share of
16:58 the actual new investment into the
17:00 market, theoretically meaning that there
17:01 would be less investor dollars for
17:03 everything else. Now, that might be a
17:06 problem for the market, not open AAI,
17:07 but at the same time, there are other
17:09 big IPOs rumored from SpaceX and
17:11 potentially even Enthropic.
17:12 Collectively, these companies could
17:14 raise more money this year than the last
17:16 decade of IPO activity combined. That
17:19 cash has to come from somewhere. And the
17:20 fear is that the market just won't be
17:22 able to bear this much supply in such a
17:24 short period of time. According to the
17:26 Z1 accounts of financial activity, the
17:29 $450 billion that just these companies
17:31 could look to raise would be more than
17:33 the net buying activity seen in US
17:35 markets in a lot of years. And that's
17:36 ignoring every single other company that
17:38 people might want to invest in. If
17:40 OpenAI can't raise this cash, it will
17:42 put more question marks over its future
17:44 and proposed spending commitments. The
17:46 sheer size of these potential new public
17:47 companies is also raising another
17:49 concern, which is that large indexes
17:51 that track the stock market will have to
17:53 readjust to include them. Index funds
17:55 from companies like Vanguard, Black
17:56 Rockck, and State Street automatically
17:58 adjust their holdings to weight their
18:00 indexes proportionally to the value of
18:02 the companies within that index. The
18:03 most valuable companies in the world
18:05 today all started out as relatively
18:07 small cap stocks that just kept on
18:09 growing, giving these indexes time to
18:10 weigh them more heavily as they
18:13 gradually grew. If these three IPOs all
18:15 go ahead at their proposed values, they
18:16 would instantly jump close to the top of
18:18 these indexes, which could force these
18:20 large asset managers to sell down a lot
18:22 of other companies in order to correctly
18:23 weight their indexes to reflect the
18:26 market. Now, not to get too nerdy, but
18:27 if only 10% of these companies are
18:29 actively sold on the market, then most
18:31 indexes are float adjusted, meaning that
18:33 only 10% of the company would be counted
18:35 towards overall holdings. But even this
18:37 could present a major shakeup in a short
18:39 period of time. My good friend Ben Felix
18:41 actually spoke directly to Vanguard's
18:42 global head of investment implementation
18:45 in a podcast he did last year. So, if
18:46 you want to get really nerdy, they go
18:48 into a lot of the technical details
18:50 about how this could play out, and I
18:51 will leave a link to that in the
18:53 description below. Now, nobody really
18:55 knows exactly how these challenges will
18:57 play out, including the company
18:59 themselves. But unfortunately, the trend
19:01 is towards a majority of this risk being
19:03 dumped onto public investors or just the
19:05 public in general. If these companies
19:07 fail, they could bring down the market
19:08 with them. But if they succeed, it will
19:10 likely be because they were used to
19:12 displace a large share of the workforce.
19:14 If the uh optimistic outcome is a
19:16 majority of human work becoming
19:18 redundant, go and watch this video next
19:19 to find out why these very same tech
19:21 bros want you to have as many children
19:23 as possible. And don't forget to like
19:24 and subscribe to keep on learning how
