Big Tech Is Faking Revenue

Massive AI Deals and Financial Circularity [0:00]

• OpenAI announced a $100 billion deal with AMD for 5 million graphics cards, with AMD returning $100 billion in stock.
• OpenAI also planned to buy $200 billion worth of graphics cards from Nvidia, with Nvidia investing $100 billion back into OpenAI.
• A deal with Oracle involved OpenAI paying $300 billion for data center usage. Oracle would then use this money to buy graphics cards from Nvidia.
• Nvidia, in turn, would use the money received from Oracle to invest back into OpenAI.

The speaker uses these large, interconnected financial announcements as a starting point to explain the concept of roundtripping. The presented deals suggest a complex flow of money between companies, raising initial questions about the authenticity of the business activity.

Did you get all that? No? Well, don't worry because in this video I am going to explain a completely unrelated concept called roundtpping.

Roundtripping Explained [1:14]

• Roundtripping is a common type of fraud where companies pass the same money in a circle.
• The purpose is to mutually inflate their valuations by pretending to do more business than they actually are.
• An example illustrates three businesses, each giving $100 to the next in a circle, resulting in each business showing $100 in revenue.
• If a company's valuation is 50 times its annual revenue, this $100 in fake revenue could make a business worth $5,000, allowing it to raise more money from investors.

This section clearly defines roundtripping and illustrates its mechanics with a simple example, highlighting how it can artificially boost valuations and attract investment.

It's like if there were three people standing in the room, right? And each one of them had a business and all of these businesses were brand new. They've never made any revenue.

Legality and Allegations of Roundtripping [2:35]

• Roundtripping is generally considered illegal in the United States and other countries because it artificially manipulates the market and inflates stock prices.
• It is a form of securities fraud.
• The speaker explicitly states they are not insinuating or suggesting that companies are engaging in illegal activities for self-enrichment, but rather analyzing the transactions.

This part addresses the legal implications of roundtripping, framing the subsequent analysis as an examination of potential patterns rather than direct accusations.

So clearly, just so there is no misunderstanding, I could not possibly insinuate or suggest that the tech bros at all of these companies would be doing something that is illegal to enrich themselves. That is 100% definitely not what's happening here.

Detailed Deal Analysis and AI Bubble Dynamics [3:05]

• Investors and Microsoft provided $58 billion to OpenAI in funding rounds.
• OpenAI then used this money to pay Microsoft for using their Azure data centers.
• Microsoft used this money to buy graphics cards from Nvidia for their Azure data centers, which OpenAI would then use to train models.
• Other companies like Google, Meta, XAI, and Amazon are also buying graphics cards from Nvidia at significantly inflated prices (4-6 times previous costs) since November 2022.
• Many of these companies have AI work that doesn't generate revenue but sounds futuristic.
• Anthropic also received billions from Amazon and then used that money to pay Amazon for AWS cloud services.
• XAI, founded by Elon Musk, sold Tesla shares to finance its operations, including buying Twitter and starting XAI.
• A conflict of interest is noted, as Tesla is a public company also operating in AI, with 100% of its valuation claimed to be AI-based, yet Musk heads both companies.
• XAI, after merging with Twitter and running low on cash, is seeking investment from Tesla shareholders.
• XAI received funding from Nvidia, which it used to buy more Nvidia graphics cards for its data center.

This section delves into specific, complex deal structures involving multiple prominent tech companies, illustrating how money and resources flow between them, often in ways that benefit hardware providers like Nvidia. The case of XAI and Tesla highlights potential conflicts of interest and the blurring lines between separate corporate entities.

The story really starts when a bunch of investors along with Microsoft find 58 billion down the back of the sofa and give it to OpenAI in various funding rounds.

Escalating Deal Sizes and Financial Strain [8:21]

• Nvidia's share price increased over 1,600% since ChatGPT's arrival.
• Nvidia announced a $100 billion investment into OpenAI, which is questionable as Nvidia lacks this amount to invest.
• OpenAI is planning to use this $100 billion investment to buy $200 billion worth of Nvidia chips.
• OpenAI recently raised $6.6 billion and is burning over $1 billion per month (before capex), potentially billions with capex.
• The announced $200 billion chip purchase from Nvidia is to be funded by a $100 billion investment from Nvidia and further funds.
• OpenAI's $300 billion deal with Oracle for data centers also involves Oracle paying Nvidia, which then reinvests in OpenAI.
• An AMD deal involves OpenAI buying $100 billion worth of graphics cards.
• Due to a lack of funds, AMD is giving OpenAI 10% of its total stock value (approximately $30 billion), allowing OpenAI to leverage this stock to borrow money for the cards.

This part details increasingly massive deal announcements, emphasizing the apparent financial strain on companies like OpenAI, which are burning through cash. The strategy of using stock of partner companies as collateral for purchases is highlighted as a sign of extreme leverage.

The company that is currently absolutely incinerating money, losing a ton of money every month and keeps doing funding rounds for relatively small amounts to keep them going, is going to buy $200 billion worth of chips from Nvidia with the $100 billion worth of investment that they're getting from Nvidia.

Leveraging Debt and Off-Balance Sheet Financing [10:43]

• XAI's funding round includes $7.5 billion in equity and up to $12.5 billion in debt through a special purpose vehicle.
• Chips purchased by XAI will be leased to the company for 5 years, allowing financiers to recoup their investment.
• This method is described as letting investors buy the graphics cards (off the books) and then lease them to the company, rather than the company using debt to buy them directly.
• This is likely structured to keep the debt off XAI's balance sheet.
• Other players like Coreweave are also involved in multi-billion dollar deals with Nvidia and OpenAI, with money moving back and forth.

This section focuses on the financial engineering used to acquire assets like graphics cards, employing debt structures and special purpose vehicles to manage financial obligations and potentially obscure liabilities.

So instead of buying the graphics cards using debt, they are letting investors buy the graphics cards because that's off the books and then lease those cards to the company for 5 years from the investors, which is probably roughly the useful life of these graphics cards anyway, running at 100% capacity for the entire duration.

Unsustainable Revenue and Profitability in AI [11:43]

• OpenAI's current revenue is around $10 billion per year.
• Anthropic's revenue is about $5 billion per year.
• Growth rates are slowing to below 100%.
• The current revenue growth is primarily from businesses paying for commercial licenses, often distributed widely without strict checks.
• Despite revenue, these companies are losing money, with losses increasing as revenue grows.
• The physical cost of producing AI answers is higher than what people are currently willing to pay for them.
• There is no foreseeable business model that can achieve profitability in the short to medium term.
• The cost of running LLMs is increasing, and user expectations for usage per dollar are rising.

This part of the transcript shifts to the financial realities of the AI companies, highlighting that despite significant revenue, they are unprofitable. The core issue is the mismatch between the cost of AI services and market willingness to pay, making current business models unsustainable.

The more revenue that they take, the more money that they're going to lose.

The "Infinite Money Glitch" and Potential Collapse [12:49]

• The massive flow of money between companies is described as "fugazi" (fake or not real), existing only on paper.
• Valuations are astronomically high due to the expectation that companies will eventually figure out how to make money, potentially through ads.
• The introduction of ads into chatbots could diminish their appeal and user trust.
• There is an ethical dilemma in recommending products based on advertising revenue rather than product merit.
• The current situation is an unprecedented example of roundtripping, with companies donating significant portions of their stock.
• This is done because the immediate boost in share price from such announcements generates tens of billions of dollars in value.
• The bubble can only unravel in two ways: either companies significantly increase their revenue and profit margins, or the "house of cards" collapses when the "infinite money glitch" runs out.

The conclusion draws a stark picture of the AI market, characterizing the current financial activity as a speculative bubble. The speaker outlines the potential scenarios for its inevitable collapse, emphasizing the current lack of a viable path to profitability.

All of this movement is just on paper. It's fugazi. Fugazi. You know that I give you $100 billion. You give it to that guy over there. That guy is going to give it to that guy over there. That guy's going to come back and give it back to me. We've all made $100 billion.