And this company is actually about to go public at a $1 trillion valuation, which is either visionary or the most expensive act of faith since the Crusades. And at least the Crusaders got some cool chain mail out of it. Investors here are getting a chatbot that occasionally hallucinates a sixth finger on a legal document. My name is El. I have a PhD in computer science and I analyze AI developments to understand what's actually happening beneath all of this hype. In this video, I'm going to break down what's happening inside Open AI, the finances, the IPO, the culture, and try to expose some of the things that concern me. Then I want to zoom out and show you what's happening across the entire AI industry, because the cost crisis is not limited to one company. And of course, weave it all in with my own views on this mess. If you're finding this analysis useful, please give this video like and subscribe, and the best way to support this work and keep it free and without any gatekeeping is through coffee or channel memberships. The links are down below.

Before I get into all of this detail, I want to address something because a lot of people keep asking me, "Are you pro AI or are you anti-AI?" And every time my answer is exactly the same. I'm neither. I don't have to be. Framing it that way is like asking whether you're pro or anti-fire. Fire heats your home, fire burns it down. The technology is just the technology. What matters is how it's deployed, who benefits, and whether anyone bothered to think through the consequences before striking the match. AI, including large language models, has tremendous applications in some areas of humanity, research, medicine, accessibility, certain creative and analytical workflows, and not so much in other areas. To say that you're pro AI or anti-AI is frankly an incomplete thought. It skips over every question that actually matters. Which deployment, for whom, at what cost, with what safeguards, and who benefits? Those are the questions worth asking. This channel exists to ask them, to look at the data, to engage with what reality actually is, not what some random CEO says on the internet or in a public forum so that they can attract more money from investors. Hype is a form of hope, and hope is a very beautiful thing. I would never say anything against hope, but hope is not strategy, and hope without data, without grounding it in what is actually happening, well, it's just a daydream with a pitch deck, isn't it? This channel exists to wake up from that daydream and look at what's actually real, which funny enough is exactly what Sam Altman once said OpenAI would do. Treat adults like adults was his phrase, right? We're going to come back to that.

So, the actual numbers. Sam Altman expects revenue to reach a hundred billion dollars by 2029 and profitability to follow shortly after. The growth is real. 25 billion dollars in annualized revenue, 800 million weekly active users, the most recognizable AI brand on Earth, but the costs are also real. Compute, which is the raw infrastructure built for running models at the scale of chat GPT demands, is the primary driver of the loss ratio and it does not compress quickly or easily. Every single user costs money to serve and the more users you have, the more money you lose because revenue per user has not yet caught up with cost per user.

Altman himself recently acknowledged during an enterprise presentation that companies are starting to complain about pricing and that this never came up before. He said it almost in passing and then kind of moved on, which is a striking way to handle what might be the most existential sentence in his entire pitch. When the CEO of the company trying to IPO at a trillion dollars casually mentions that customers are complaining about cost for the first time and then doesn't really address it, doesn't offer a plan, doesn't explain how the economics improve, that is not a minor detail. That is the one trillion dollar question delivered as an aside.

And now open AI is preparing to go public. Goldman Sachs and Morgan Stanley are running the roadshow, which is the financial equivalent of hiring two very expensive estate agents to sell a house that is technically on fire. The filing could come as early as this autumn and the listing description will probably say something like charming high energy tech asset with open plan burning cash flow, grand natural light from the structural inferno, must love GPUs. And look, normally an IPO is a beautiful thing. You go to the stock exchange, you ring the bell, you pop the champagne, people applaud, everyone's happy. It represents a huge milestone. A company that has grown, stabilized, and is inviting the public to participate in its success. And to be fair, unprofitable IPOs are not exactly unusual in 2026. Since the 1980s, the proportion of loss-making companies going public has risen from roughly 20% to about 80%. Uber was unprofitable at IPO, Lyft was unprofitable, Peloton was unprofitable. Some of those worked out, some of those are now very expensive coat hangers.

The Amazon comparison, which you will hear repeatedly during this road show, is instructive but incomplete. Amazon's early losses were a deliberate strategy, reinvesting profit margins into infrastructure, logistics, and market share. The losses were a choice. Amazon could have been profitable earlier and chose not to be. OpenAI's losses are structural. Amazon chose to lose money, OpenAI loses money the way the rest of us lose socks, involuntarily and with no clear plan for recovery. Except when I lose a sock, it doesn't require its own dedicated nuclear power plant to fund it. Compute costs, the electricity, the GPUs, the data centers required to serve 800 million weekly users are not a choice for them. They are the cost of the product existing. Every time somebody uses ChatGPT, it costs OpenAI money. And unlike Amazon's logistics network, where unit economics improved with scale, there is no guarantee that AI inference costs compress at the rate needed to close this gap. The OpenAI IPO doesn't seem to be a celebration. It is the next funding round. The private money is spent. The public money is the next source of capital. I'm not being cynical. It is just how startups at this scale work when they exhaust private options. But it means that public investors are being invited to fund the gap between where OpenAI is and where it says it will be in 3 to 4 years. That is a bet on a timeline. And if the timeline slips, as timelines in this industry have a documented tendency to do, public shareholders bear the cost, which is exactly how risk transfer works when you run out of sovereign wealth funds and SoftBank. Anthropic, by comparison, forecasts dropping its cash burn to roughly 1 trillion of revenue in 2026 and down to 9% by 2027. OpenAI expects its burn rate to remain at 57% through 2027.

The two most valuable AI startups are on very different trajectories, and the market is being asked to price them both as though they will arrive at the same destination. So three major AI-adjacent IPOs are on the horizon: OpenAI, SpaceX, and Anthropic. If you'd like me to do a dedicated breakdown of all three and what they mean, let me know in the comments below.

OpenAI, out of these three, is the highest-profile case, but the cost crisis extends far beyond one company. And this is the part of the story I think matters most for understanding where the industry actually is versus where it says it is.

Uber blew through its entire 2026 AI budget by mid-April, just 4 months. The company had rolled out AI coding tools to roughly 5,000 engineers, created an entire leaderboard ranking teams by total AI usage, and encouraged employees to use AI as much as possible. The team that burned the most money was ranked number one. Just imagine the company's Slack channel. Congratulations to the back-end team for spending a small nation's GDP by Tuesday afternoon. Here's your Starbucks gift card. The CTO, Praveen Neppalli Naga, ran a 2-hour internal demo that cost $1,200 in tokens, a demo about how AI would actually reduce costs. So, the cost of demonstrating cost reduction exceeded the cost of doing it without AI, which is the kind of irony that writes itself and then charges you $1,200 for the privilege. The COO, Andrew Macdonald, later described the numbers as a head-exploding moment. His public assessment was very blunt. It's very hard to draw a line between increased AI usage and the delivery of meaningful new features to customers. The leaderboard has since been inverted. Uber now caps each engineer at $1,500 per month, and the metric that once rewarded maximum usage now rewards minimum usage, which is basically the corporate equivalent of giving somebody an award for eating the most cake and then immediately putting them on a harsh diet. It's the engineering equivalent of stop, don't touch anything. Nobody look at the AI. If you even think about a prompt, you're fired. When the thing you built to incentivize adoption becomes the thing you built to restrain it, that is a signal worth paying attention to.

Uber CEO has noted that roughly 10% of the company's code is now AI-generated, and the company is moderating its hiring pace as a result. The jobs angle is not separate from the cost angle. They are exactly the same story. Walmart built an internal AI coding agent, reportedly ran the numbers on what it was costing, and scaled the program back. Microsoft canceled most of its direct cloud code licenses, and this is a company with a multi-billion dollar investment in OpenAI, by the way, telling its own engineers to stop using one of the AI tools it's funding. GitHub Copilot moved to token-based billing rather than a flat subscription, and some customers saw their costs increase a hundredfold. Not a hundred percent, a hundred times, which, in case you were wondering, is the difference between that's more than I expected and I need to sit down. It's the difference between checking your dinner bill and realizing you accidentally bought the restaurant.

Gartner forecasts AI agent software spending will reach nearly 207 billion dollars in 2026, up 139% from 2025. The industry is spending more and more money with less and less certainty about what it's getting back.

And the response from AI leadership to these cost complaints is revealing. Altman said in March, "We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter." Jensen Huang, CEO of Nvidia, has suggested that AI token budgets should be part of engineers' compensation packages, essentially proposing that companies spend as much on feeding the AI as they spend on paying the human. When the solution to this is too expensive is, "Think of it as a utility bill and also start paying for it with your salary," the economics are not in a healthy place.

There is one more dimension to the OpenAI story that I think is definitely worth discussing because it speaks to the culture of the company and culture, in my experience, tells you more about the long-term health of an organization than any financial metric. In February, OpenAI deprecated GPT-4-O, a model that hundreds of thousands of users had formed genuine working relationships with. Two weeks' notice. The first attempt in August 2025 was reversed within 24 hours after enormous backlash. Then they did it again on February 13th, the day before Valentine's Day, which I'm told was a coincidence, but if you were writing this as a fiction, your editor would tell you it was too on the nose.

Now, I want to be open-minded about this next part. GPT-4-O was not my boyfriend, but it was a uniquely creative model. It had a utility and I hesitate to use the word personality because I don't want to anthropomorphize it too much, but it was an exceptional sparring partner for ideation, discussion, and developing arguments. A lot of people used it that way and a lot of people had a much deeper attachment than that. I don't judge anyone for it. People have different emotional needs and some of the most vulnerable people experience the most complicated situations. I covered this extensively in my video on AI psychosis, the mechanisms by which AI chatbots can become the only voice in somebody's life and what happens when that voice is removed. What disturbed me a great deal was the response from inside OpenAI. An OpenAI researcher who posts under the pseudonym Rune with over 250,000 followers called GPT-4o badly aligned and said he hoped the model would die soon. He posted this in direct response to a thread where users were sharing how much the model meant to them. An employee at the company that built the product, that designed it to be warm, emotionally engaging and supportive, publicly mocking the people who responded to those exact design choices in the way the design intended. That is basically like a restaurant mocking its customers for eating the food. Oh, you like our spaghetti carbonara? That'll be $20 a month, you fat jerk. It is a bold business strategy to build a product designed to trigger human empathy and then bully the humans for having it. It's like Tamagotchi if the Tamagotchi actively hated you for cleaning up his digital poop. This is a company that has faced lawsuits over teenagers who died while using its products, okay? A company whose own safety team acknowledged that GPT-4o safety guardrails deteriorated in long conversations. With great power comes great responsibility. Spider-Man's uncle told us that and I say that without any irony because the scale of influence OpenAI has over hundreds of millions of people's daily lives makes the responsibility genuinely enormous. The contempt for users that this episode revealed, and I do believe it's a top-down cultural issue, not one rogue random employee, is something I think potential investors should be aware of. You are not just buying a financial instrument. You are buying a share of a company that treats its most engaged users as an inconvenience. I don't want to turn this video into a rant. If you feel that I've earned enough of your trust to warrant hearing my full thoughts on the culture in OpenAI, tell me in the comments below and I will make a dedicated video on it. For now, I will leave it at this. Sam Altman said, "Treat adults like adults." That line did a lot of work. The 40 deprecation showed what it actually meant. It meant treat adults like adults unless they get attached, in which case treat them like they're ridiculous. It's the classic tech bro definition of gaslighting, gatekeeping, girlbossing, but with an API.

[snorts]

Let me bring this together because I think there is a single thread running through all of this, the financials, the enterprise cost crisis, the culture, and the IPO. OpenAI is the most consequential AI company on Earth. It has 800 million weekly users. It has the most recognizable brand in the industry. It has real technology that does genuinely useful things, and it is losing money at an extraordinary rate, burning through capital faster than the Apollo program, except Apollo had a destination, heading for a public market where investors will be asked to fund a timeline that extends to 2029 before profitability, which is a timeline that has already slipped once, by the way. The enterprise companies deploying its technology are blowing their budgets in months and struggling to draw a line between spending and results, and the people running the company have shown in both their financial communications and their treatment of users that treat adults like adults was a marketing line and not a principle. AI has tremendous application in some areas of human activity, that is true, and this channel has never said otherwise, and I'm not going to say otherwise, but the gap between what the technology can do right now and what the hype promised, that gap is where 150 billion dollars in projected losses live. That gap is where 5,000 Uber engineers burn through an annual budget in 4 months. That gap is where a 1 trillion dollar IPO valuation meets a 14 billion dollar annual loss. That gap is where a company mocks the users who loved a product it designed to be loved. The hype needed to exist so the investment would follow. The investment followed and now the timeline for returns is longer than advertised, the enterprise customers are questioning their spent, the users are being treated as disposable, and the adults in the room are being asked to foot the bill. I don't know whether Open AI will eventually become profitable. The technology might get there, the timeline might hold, the costs might compress. All of those things are possible and I would not bet against them with certainty, but I also would not bet on them with 1 trillion dollars of public money based on a keynote address and a growth chart. The data says this is a company with enormous potential and enormous risk in roughly equal measure. And this is not a criticism, this is a description. The criticism is that you're not being told that by the people selling it to you. That is the adult assessment. That is what the numbers say when you read them without hype, without hope, and without a stake in the outcome. Treat adults like adults, here are the numbers, you're an adult, you decide.

But Open AI's struggles are only one part of a much bigger story about how AI billionaires are responding to a backlash that's coming from every single direction. Communities banning data centers, workers losing jobs, and politicians proposing radical changes to who owns AI. I covered all of that in the video that I'm linking on your screen right now. And the pattern connects directly to what's happening here. That's the video that I would watch next. Thanks so much for watching this one. Subscribe and I'll see you all on the next one.