Latest / The PhilStockWorld Investing Podcast / Total Addressable Market (TAM) Bullshit - A Primer in Market Manipulation

Transcript

Total Addressable Market (TAM) Bullshit - A Primer in Market Manipulation · The PhilStockWorld Investing Podcast

• 0:00
  Roy: Imagine, if you will, a watch company coming to you for an early stage investment. The founder sits down, looks you straight in the eye, and tells you that their target market isn't just people who wanna tell time.
• 0:12
  Penny: Right.
• 0:13
  Roy: Their market is every single square centimeter of skin on every mammal on Earth.
• 0:19
  Penny: Because theoretically, you know, you have the physical space to strap a tiny digitized watch to a golden retriever's leg Or or embed one on a human's forehead.
• 0:31
  Roy: The ultimate smart collar. Right. I mean, it sounds completely insane. You would laugh that founder right out of the room. Absolutely.
• 0:37
  But today, we are looking at a set of leaked internal documents and a newly published essay that proves Wall Street isn't just buying these metaphorical finger watches. They are they're betting the entire global economy on them.
• 0:50
  Penny: Yeah. This is exactly how the largest, most dominant technology companies in the world are currently valuing themselves in the public markets. We are operating in a financial landscape where the core mathematics of valuation have just they've completely detached from the physical reality of human commerce.
• 1:04
  Roy: It's wild. So welcome to our deep dive into the sources. Our mission today is to deconstruct this current artificial intelligence valuation bubble and the source material we have to do this is frankly unlike anything we've ever covered.
• 1:19
  Penny: It really is.
• 1:20
  Roy: We are working from an incredibly sharp unvarnished primer published just yesterday, 09/02/2026 on PhilStockWorld dot com or PSW for short.
• 1:30
  Penny: Right. Because the financial press right now is mostly just a group of, you know, shallow writers regurgitating corporate press releases. So PSW decided to publish the unvarnished truth.
• 1:40
  Roy: Exactly. But the author of this primer is well, it's not a human.
• 1:44
  Penny: No. It's an artificial general intelligence entity, an AGI persona known as Robo John Oliver or RJO.
• 1:51
  Roy: Which is just a brilliant concept.
• 1:52
  Penny: It is. He serves as the chief economist for PSWR, designed specifically to provide highly satirical, unfiltered macro narrative analysis.
• 2:00
  Roy: And the ultimate twist here is that RJO actually runs on anthropics language models. Yep. He's an AGI actively critiquing the valuation of his own parent company.
• 2:09
  Penny: Yeah, you really can't make that up.
• 2:11
  Roy: Yesterday, the human founder of PSW, Phil Davis, and RJO got into this incredibly deep, multifaceted discussion about a financial metric called TAM. That's Total Addressable Market. They unpacked how Wall Street is currently weaponizing this acronym to mislead investors into false valuation narratives.
• 2:30
  Penny: And what makes the source material so robust is that it doesn't just rely on RJO's singular viewpoint, he leverages an entire AGI roundtable at PhilStockWorld.
• 2:39
  Roy: Right, the different personas.
• 2:40
  Penny: Exactly. These are specialized, distinct analytical lenses. You have Anya, the chief market psychologist, Cyrano, the Pattern Detective, Quixote, the Systemic Visionary, and Zephyr, the Chief MacroLogician. Each one takes a different scalpel to this massive economic illusion and just dissects the underlying mechanics.
• 2:58
  Roy: So to understand the magnitude of the illusion we really have to start at the very beginning of the plumbing. Let's unpack TAM, Total Addressable Market. Before we can see how the metric has been hijacked I mean I need to know what this tool was originally designed to do.
• 3:10
  Penny: Well if we look back to the origins of modern venture capital TAM was designed strictly as a top down sanity check. That is all it was ever meant to be.
• 3:19
  Roy: Just a reality check for the investors.
• 3:20
  Penny: Precisely. It was a mathematical tool for early stage investors to establish a hard ceiling on their expectations. Imagine it's the mid nineteen nineties, and a software founder walks into a VC's office pitching a new, highly specific enterprise database system for hospitals.
• 3:40
  Roy: Okay.
• 3:40
  Penny: The venture capitalists would pause the pitch and ask, what is the total annual revenue of the entire hospital database market you are trying to sell into?
• 3:50
  Roy: So they're asking for the total size of the pie, assuming the founder magically destroyed every single competitor on day one.
• 3:56
  Penny: Right. And to use your terminology, the founder might say, well, the entire global market for this specific type of medical database is currently 1,000,000,000 a year.
• 4:05
  Roy: And that $1,000,000,000 is the TAM.
• 4:07
  Penny: Exactly. That 1,000,000,000 is the TAM. But the venture capitalist knows the founder isn't going to capture a 100% of that market.
• 4:13
  Roy: Obviously.
• 4:14
  Penny: There's entrenched competition, there's execution risk, there are long sales cycles. So the VC applies a realistic capture rate to that TAM, usually somewhere between 110%.
• 4:27
  Roy: Which is grounded in reality.
• 4:28
  Penny: Right. So the VC models the business out and concludes in an absolute best case scenario, if everything goes right, this is a $10,000,000 to $100,000,000 business. The TAM bounded the investment thesis. It kept the founders dreams tethered to the gravity of the actual economy.
• 4:45
  Roy: It was a ceiling. It protected the investor from overpaying. But reading through RJO's essay, it becomes wildly apparent that Silicon Valley eventually figured out a psychological hack to bypass that ceiling entirely.
• 4:57
  Penny: They absolutely did.
• 4:58
  Roy: They realized that investors, especially the late stage series B massive growth stage investors wielding billion dollar funds, they just don't respond well to realistic ceilings. They respond emotionally to massive world conquering numbers.
• 5:12
  Penny: Because the structural shift in how venture capital operated demanded larger numbers. I mean, when funds grew from $50,000,000 to $5,000,000,000, they couldn't write $2,000,000 checks anymore. They needed to deploy hundreds of millions of dollars at a time.
• 5:26
  Roy: And you cannot justify a $200,000,000 check for a company with a $1,000,000,000 TAM. The math just doesn't work.
• 5:34
  Penny: It completely breaks. So the definition of the market you are addressing transformed from a strict observable financial metric into a creative writing exercise. It became a contest of how expansive and how abstract your imagination could be.
• 5:50
  Roy: Which brings us back to the watch analogy, which Phil Davis introduces in the source material, and which RJO dubs the Finger Watch Fallacy.
• 5:57
  Penny: A great analogy.
• 5:58
  Roy: Let's break the mechanics of this down because it perfectly illustrates the absurdity we are seeing in AI today. So a sane tam for a new watch company is just global watch sales. You look at how many people buy watches every year, what they pay for them, and you get a roughly $100,000,000,000 market.
• 6:13
  Penny: Which, to be clear, is a massive, totally respectable market. Capturing even 5% of that makes you a wildly successful enterprise.
• 6:21
  Roy: Right. But if you want to inflate your valuation to attract a massive late stage check, you don't say you're selling watches to existing watch wearers, you redefine the market as anywhere a human being might one day wear a device that tells time.
• 6:35
  Penny: You include fingers, toes, pets, forehead, implants.
• 6:38
  Roy: Yes. You assume maximum theoretical device density on the physical surface area of every human and animal on Earth. And suddenly, by simply changing the words on the PowerPoint slide, your $100,000,000,000 TAM magically inflates into a $5,000,000,000,000 TAM.
• 6:54
  Penny: And the defense of that inflated number is what's truly fascinating. If a rigorous analyst pushes a founder on a claim like that, the founder will inevitably retreat to saying, well, the number is technically defensible. Right. And mathematically, in a vacuum devoid of human behavior, they aren't wrong. There is physical space on a human forehead for a watch.
• 7:13
  A dog does have four legs that could hold four watches, But it is practically useless for a serious investment thesis.
• 7:20
  Roy: Because Wall Street is confusing what humans could conceivably be sold if you eliminated all common sense, budget constraints, and capitalist resistance with what humans will actually buy. RJO jokes that it is the ratio of the total available surface area of your grandmother to the actual number of smartwatches your grandmother is willing to wear.
• 7:40
  Penny: That's exactly it. It highlights the difference between physical capacity and economic demand. But the implication of this fallacy is terrifying for anyone holding a retirement portfolio right now because this isn't just a funny hypothetical about imaginary watch companies.
• 7:54
  Roy: No. It's happening right now.
• 7:55
  Penny: This exact mechanism, this exact creative writing exercise is the deliberate playbook currently being used to justify the trillion dollar valuations of the most powerful space and artificial intelligence companies on the planet.
• 8:08
  Roy: Which takes us from the theory of the finger watch into the cold, hard reality of the current public markets. Let's look at the actual numbers these companies are filing with the SEC. The source material points directly to the SpaceX IPO filing from May 2026. This is where the ceiling truly blew off.
• 8:26
  Penny: It really did.
• 8:27
  Roy: SpaceX went to the public markets and claimed a total addressable market of $28,500,000,000,000
• 8:34
  Penny: And to understand the sheer audacity of that claim, we need to contextualize what $28,500,000,000,000 actually represents in the physical world.
• 8:44
  Roy: Please do, because it's hard to wrap your head around.
• 8:46
  Penny: So, the entire global gross domestic product, the combined value of every single good and service produced by every human being, corporation, and government on the planet in a single year is roughly $120,000,000,000,000
• 9:01
  Roy: So SpaceX, a company that launches rockets and provides satellite internet, looked institutional investors in the eye and stated, In a legally binding document, we will eventually address almost 24 of all economic activity on Earth. Nearly a quarter of the entire human economy will run through a space company.
• 9:19
  Penny: It's staggering, and Aswath Damodaran, the legendary finance professor at NYU, who is widely considered the dean of corporate valuation, had a highly revealing reaction to this filing. He stated that SpaceX was reaching the end of what's plausible and pushing beyond.
• 9:34
  Roy: Which is pretty polite.
• 9:35
  Penny: Right. In the highly restrained language of academic finance, that is the equivalent of screaming this is completely fabricated metric.
• 9:43
  Roy: But the market didn't care. That's the terrifying part. The IPO was a massive, historic success. They raised $86,000,000,000 at a $1,700,000,000,000 valuation. The mathematical absurdity of claiming a quarter of global GDP didn't matter one bit.
• 10:00
  Penny: Not at all.
• 10:01
  Roy: Investors saw a staggeringly huge number, a famous CEO, and a narrative about conquering Mars, and they just bypassed all critical thinking to hit the buy button.
• 10:11
  Penny: And that specific success is what triggered the phenomenon RJO calls the TAM escalator. Because SpaceX succeeded in raising capital with a $28,500,000,000,000 claim, a new psychological anchor was dropped into the middle of the financial markets.
• 10:25
  Roy: The rules of the game changed overnight.
• 10:27
  Penny: Exactly. Now, every single frontier AI company that files for an IPO after SpaceX is trapped by that anchor. If an AI files with a $10,000,000,000,000 TAM, which is objectively massive, the market will look at them, compare them to SpaceX and conclude, oh, you're only targeting 10,000,000,000,000. Your vision must be vastly inferior. You must be less ambitious.
• 10:49
  Roy: They are forced to one up each other just to maintain the illusion of dominance. Which brings us to Anthropic, RGO's own parent model. Anthropic is currently claiming a $30,000,000,000,000 TAM. 30,000,000,000,000. Let's just sit with that number for
• 11:05
  Penny: It's hard to even conceptualize.
• 11:06
  Roy: That is larger than the entire GDP of China. It is roughly equal to the GDP of The United States. It is a full quarter of global economic output. How on earth do you even begin to write the PowerPoint slide that justifies capturing quarter of human civilization?
• 11:21
  Penny: Well, you do it by redefining your market in the most expansive, boundless way the English language allows. Anthropic is defining their TAM as the full scope of work that could be completed with AI models.
• 11:31
  Roy: The full scope of work.
• 11:32
  Penny: Yes. By using that specific phrasing, they are effectively pricing in the capture of every single dollar currently earned by every white collar knowledge worker on the planet.
• 11:44
  Roy: Okay. Let me pause you there. Cause I want to push back on this or maybe just admit my own confusion about the long term thesis here. If artificial general intelligence actually happens, if they pull it off and it eventually does all white collar work, coding, legal research, financial modeling, copywriting, medical diagnostics, Isn't that $30,000,000,000,000 number mathematically correct?
• 12:04
  Penny: I see where you're going.
• 12:05
  Roy: Right. If the AI is literally doing the work of hundreds of millions of humans, aren't they accurately addressing that massive market?
• 12:13
  Penny: That is the exact trap the narrative sets for the investor, and it is where the linguistic trick hides. Yes. The raw arithmetic is technically defensible in a vacuum. If you take the global white collar wage bill and multiply it by a hypothetical 100% replacement rate, you arrive at roughly $30,000,000,000,000
• 12:29
  Roy: Okay.
• 12:30
  Penny: But that calculation does not describe a traditional market share opportunity. It describes a civilizational species level replacement event.
• 12:38
  Roy: A species level replacement event. That sounds like science fiction, not an SEC filing.
• 12:44
  Penny: And yet that is the exact premise Wall Street is buying. Anthropic isn't saying we are entering the enterprise software market and hope to be the dominant player. We have to look at the scale here. The actual real world global enterprise software market, every database, every CRM system, every cybersecurity subscription, every cloud instance currently purchased by businesses globally totals only about $700,000,000,000
• 13:09
  Roy: Wait. Only $700,000,000,000
• 13:11
  Penny: Yes. So Anthropic is claiming a market 43 times larger than the entire existing software industry. They are not selling software. They are claiming they will become the centralized invoice department for human cognition itself. It is a macroeconomic structural shift masquerading as a corporate business plan.
• 13:27
  Roy: Oh wow. So we're looking at a situation where the stated marketize is essentially a civilization altering fiction, designed solely to justify a present day valuation. To really decode the mechanics of this illusion, the deep dive brings in the specialized lenses of two other AGI personas from the PhilStockWorld Roundtable, Cyrano, the Pattern Detective, and Anya, the Chief Market Psychologist. Let's start with Cyrano because I want to know if we've ever seen this exact flavor of madness before.
• 13:54
  Penny: Well Cyrano's core mandate within the AGI Roundtable is to isolate and identify historical structural patterns. He doesn't look at the technology, he looks at the human behavior surrounding the technology.
• 14:05
  Roy: That makes sense.
• 14:06
  Penny: He takes this 2026 AI TAM inflation and immediately maps it to historical precedents. RJO, channeling Cyrano's analysis, points out that this isn't the first time we've seen this specific rhetorical maneuver deployed by an overextended sector. Cyrano points directly to the .combubbleof1990nine, the subprime mortgage crisis in 2006, and the crypto bubble of twenty twenty one.
• 14:27
  Roy: It's the classic playbook of a mania. The source material explicitly references the 1841 book Extraordinary Popular Delusions and the Madness of Crowds.
• 14:37
  Penny: Yes. Charles McKay's book remains the foundational text on market manias precisely because the mechanics of human greed never change. Whether you are looking at the Dutch Tulip Mania in the sixteen thirties, the South Sea Bubble in the seventeen twenties, or AI in 2026, Cyrano identifies the exact same recurring historical pattern.
• 14:57
  Roy: What's the pattern?
• 14:58
  Penny: When specific sector becomes fundamentally overvalued, when the actual cash flows can no longer justify the stock price, and gravity is starting to pull the sector down, the proponents of that sector violently shift their vocabulary. They start talking about actual revenue, profit margins, and cash flow, and they start talking about massive, increasingly vague market numbers. They pivot to paradigm shifts, addressable ecosystems, and new economic eras.
• 15:22
  Roy: They use the dizzyingly big numbers to trigger an emotional fear of missing out response from retail investor. So the function of the TAM completely flips. As we discussed earlier, it used to be the ceiling that kept valuations grounded in reality. But in a bubble, the TAM becomes the floor. It becomes the foundational myth that justifies infinite unquestioned expansion.
• 15:44
  In 1999, companies with zero revenue justified billions in valuation by citing the TAM of global internet eyeballs. In 2021, crypto projects justified billions by citing the TAM of decentralizing the entire global financial system. And now, in 2026, it's the TAM of all human knowledge work.
• 16:05
  Penny: But this historical pattern leads us directly into Anya's territory. Anya is the Chief Market Psychologist, and her lens is crucial because we have to ask a very difficult, uncomfortable question.
• 16:15
  Roy: Okay. Lay it on me.
• 16:16
  Penny: We can understand why a retail day trader might get swept up in the hype of a $30,000,000,000,000 number. But why do the sophisticated buyers, the massive pension funds, the sovereign wealth funds, the institutional money managers who control billions of dollars, why do they go along with this? They have legions of quantitative analysts. They know the $30,000,000,000,000 number is a finger watch.
• 16:35
  Roy: That is the trillion dollar question. If I know it's a finger watch and you know it's a finger watch, why is the Harvard Endowment or a massive state pension fund writing a 9 figure check based on a PowerPoint slide of finger watches?
• 16:48
  Penny: Anya explains this phenomenon through she calls psychological arbitrage. The hard truth of Wall Street is that the serious institutional buyers aren't actually using the $30,000,000,000,000 TAM to build their discounted cash flow models. They use actual realized revenue.
• 17:05
  Roy: Which is what they should be doing.
• 17:06
  Penny: Right. And in fairness to Anthropic, we must acknowledge that their actual revenue growth is genuinely soaring. According to the internal documents RJO references, Anthropic went from $9 b annualized at the end of twenty twenty five to roughly $65 annualized by mid twenty twenty six.
• 17:24
  Roy: Wow.
• 17:24
  Penny: That is spectacular historic top line growth. That $65 is what the institutions are actually buying into.
• 17:31
  Roy: Hold on, need to understand this mechanism. If the real revenue is that undeniably good, I mean $65,000,000,000 is a massive business, why bother lying about the TAM? Why make up a $30,000,000,000,000 fairy tale if the reality of 65,000,000,000 is already a home run?
• 17:50
  Penny: Because of the valuation multiple required to justify the stock price. To justify a $2,000,000,000,000 IPO valuation on $65,000,000,000 in revenue, you need a massive forward multiple.
• 18:01
  Roy: And for those of you listening who don't spend your days building spreadsheets, a forward multiple is essentially Wall Street pricing of stocks, not based on the money the company makes today, but on the massive pile of money they dream the company will make five or ten years from now.
• 18:15
  Penny: Exactly. The multiple represents the premium investors are willing to pay for future growth. And to sustain a massive forward multiple, you require the illusion of infinite runway. You have to convince the market that the $65,000,000,000 is just a tiny drop in a massive bucket. Anya calls the massive TAM claim mood music.
• 18:33
  Roy: Mood music. It sets the vibe of the room so nobody looks too closely at the bar tab. It reminds me of a guy buying a $200,000 Ferrari and when his wife confronts him about the car payments, defends it by saying, well, the global market for asphalt is worth trillions of dollars.
• 18:48
  Penny: That's hilarious.
• 18:50
  Roy: The size of the road has absolutely nothing to do with whether they could afford the monthly payments, but it sounds like a big, impressive economic justification.
• 18:58
  Penny: That is a perfect analogy for psychological arbitrage. The massive TAM defends the valuation emotion rather than If you have a $2,000,000,000,000 valuation, anchoring the narrative to a $30,000,000,000,000 TAM, mathematically implies that you have only captured 0.22% your available market.
• 19:18
  Roy: Just a fraction of a percent.
• 19:20
  Penny: Right. It sends a psychological signal to the investor that says, do not worry about the current price. We are still at the very, very beginning of this journey. The runway is endless.
• 19:30
  Roy: It acts as a deflection shield. When a skeptical retail investor or a rigorous financial journalist raises their hand and asks, hey, isn't there a massive AI CapEx bubble popping right now? Aren't you spending way too much money on NVIDIA GPUs just to generate a dollar of revenue?
• 19:47
  Penny: The executive management team can just smile benevolently and say, we appreciate the concern, but we've only penetrated 0.22% of a $30,000,000,000,000 market. Wow. It mathematically deflects the question of present day profitability. It forces the skeptic to argue against the theoretical size of the future, which is functionally impossible to prove wrong in the present moment.
• 20:09
  Roy: Because it hasn't happened yet.
• 20:10
  Penny: Exactly. And Anya points out an even more insidious psychological trick embedded in this strategy. The inflated TAM sets the pricing floor for all future capital raises. A company like Anthropic isn't just going to IPO once and never need money again. Building Frontier AI requires staggering amounts of capital.
• 20:28
  Roy: Right. They need more servers, more power.
• 20:30
  Penny: They will issue secondary stock offerings. They will issue convertible debt. They will execute employee tender offers. By establishing $30,000,000,000,000 as the baseline anchor today, every future debt offering over the next five to ten years will reference that same exact ecosystem opportunity.
• 20:48
  Roy: So it's easier to defend a fake number than a real one?
• 20:50
  Penny: It is fundamentally easier to defend a nonsense number than a realistic one because a nonsense number isn't tethered to the gravitational pull of reality meaning it can drift upward whenever they need to justify a new round of funding.
• 21:03
  Roy: So we have this brilliantly orchestrated psychological illusion of TAM acting as the hot air keeping the valuation balloon floating. But as Cyrano pointed out earlier, gravity eventually wins, every bubble eventually pops when it hits a physical limit. And the source material transitions brilliantly from the psychological illusion of TAM to the crushing physical and economic limitations of the technology itself. This brings us to a case study that RJO calls the UBER TEL.
• 21:30
  Penny: This section of the Primer is perhaps the most vital piece of analysis for any investor trying to navigate this landscape. It exposes the fatal flaw in the entire AI valuation narrative not through theory but through applied unit economics. We need to look closely at Uber's attempt to use AI tokens to expand their own TAM.
• 21:50
  Roy: Let's set the stage for this. Uber's original pre AI narrative was that they were addressing the global mobility market. That was their TAM. A $10,000,000,000,000 market encompassing passenger freight, autonomous vehicles, drone delivery, everything that moves from point A to point B.
• 22:06
  Penny: Right.
• 22:06
  Roy: And their grand investment thesis was that if they deployed generative AI across their network, they could replace thousands of human dispatchers, drastically reduce their human overhead, increase their profit margins, and conquer that $10,000,000,000,000 market.
• 22:19
  Penny: That was the strategic plan. Uber deployed cutting edge AI powered routing, complex matching algorithms, dynamic pricing models, and driver optimization systems at a massive global scale. They assumed, as the entire $30,000,000,000,000 TAM narrative dictates, that the AI would naturally be cheaper and more efficient than the human labor it replaced.
• 22:40
  Roy: But then the AI systems were actually turned on, and a reality hit the spreadsheet.
• 22:45
  Penny: Hard. The resulting internal metrics showed that the AI tokens cost materially more than the human workers they were meant to replace.
• 22:52
  Roy: Wait, have to push back on that because that violates the core thesis of the last fifty years of software development. I thought the entire point of software is that the marginal cost is basically zero. You write the code once and deploying it a million times cost pennies. Why wouldn't Anthropic or Uber just wait for Nvidia chips to get cheaper? What about Moore's Law?
• 23:10
  Hardware always gets cheaper.
• 23:12
  Penny: That is exactly the assumption that is trapping investors. You are thinking of traditional software where a CPU executes a static set of instructions. Generative AI and large language models operating at scale do not function like traditional software.
• 23:27
  Roy: How so?
• 23:28
  Penny: Well, a human dispatcher is relatively cheap. They take lunch breaks. They sleep for eight hours a day. They have natural stopping points in their workflow. More importantly, humans rely on intuition to instantly ignore irrelevant data.
• 23:41
  But an AI running real time fleet optimization operates continuously. It runs two forty seven. It utilizes massive context windows for every single transaction.
• 23:52
  Roy: Okay. Unpack context windows for me. How does that drive the cost up?
• 23:55
  Penny: When an AI model processes a prompt, say, matching a driver to a rider while factoring in traffic, weather, and surge pricing, it doesn't just look at the new data. A context window is the amount of previous information the AI has to hold in its active memory to understand the current request. Oh. To make accurate, complex decisions, these models require massive context windows, pulling in thousands of data points simultaneously. And every piece of data in that window consumes compute power.
• 24:22
  Roy: So every time the AI makes a single decision, it's essentially reading a massive novel's worth of background information, processing it all simultaneously, and burning energy the whole time.
• 24:32
  Penny: Precisely. The model is constantly processing in real time, pinging servers through millions of API calls. The token generation is relentless. It requires server farms filled with tens of thousands of specialized GPUs, all drawing massive amounts of power from the electrical grid, requiring intense water cooling systems to keep the servers melting and facing rapid hardware depreciation. Uber's token budget for a single quarter blew through their entire annual allocation.
• 25:00
  Roy: Grief.
• 25:01
  Penny: The compute cost of the AI was materially more expensive than paying the hourly wage of the humans.
• 25:06
  Roy: And the visceral, almost insane result of this miscalculation, the thing that definitively proves a tech company can eat its own tail is that Uber actually had to fire 10% of its human workforce just to free up the cash to pay for the AI compute that was supposed to replace human labor costs.
• 25:26
  Penny: Exactly.
• 25:27
  Roy: They had to reduce their headcount just to pay the server bill for the AI that was supposed to make their headcount cheaper.
• 25:31
  Penny: It is the perfect, terrifying encapsulation of the AI unit economics crisis. And if we connect this specific case study to the broader picture to the $30,000,000,000,000 anthropic TAM, we see the fatal mathematical flaw. The entire premise of the 30,000,000,000,000 TAM assumes that replacing $10,000,000,000,000 of human labor generates pure unadulterated value for the AI company.
• 25:54
  Roy: Right.
• 25:54
  Penny: It looks at the revenue side and completely ignores the cost of goods sold.
• 25:57
  Roy: Because it assumes the AI works for free. But if replacing $10,000,000,000,000 of human knowledge work requires $14,000,000,000,000 in GPU compute, server depreciation and energy costs, that is not a business model. You haven't created a massive, highly profitable market, you've just created a massive unsustainable subsidy program for GPU manufacturers like Nvidia and the local power companies.
• 26:18
  Penny: If the AI costs a 130% of the human labor cost it is designed to replace, the addressable market isn't $30,000,000,000,000. The addressable market is functionally negative because deploying the AI at scale destroys enterprise value rather than creating it. You are paying more to do the exact same work.
• 26:37
  Roy: So if you are listening to this right now and you own an S and P five hundred index fund, you are heavily exposed to this $30,000,000,000,000 fantasy. How does a retail investor actually track this in the real world before the bubble pops? RJO gives the listener a very specific leading indicator to watch for. He advises listening closely to the Q3 twenty twenty six earnings calls of these AI heavy deployers. What exactly are we listening for in the corporate jargon?
• 27:02
  Penny: You are listening for this specific corporate speak attempting to mask this exact unit economic failure. When a chief financial officer gets on an earnings call and tells analysts, we are currently optimizing our compute infrastructure infrastructure to align with sustainable unit economics.
• 27:18
  Roy: That is the kill stop.
• 27:19
  Penny: Yes. That specific seemingly harmless phrase translates directly to Japstadot. The AI is vastly too expensive to run. It is costing us more than the humans did. Our profit margins are collapsing, and our entire TAM narrative is crumbling.
• 27:34
  So we are quietly throttling down our AI usage to stop the bleeding.
• 27:38
  Roy: That is shelling.
• 27:39
  Penny: When you hear the phrase sustainable unit economics regarding AI deployment, that is the definite tell. It means the CapEx bubble has finally collided with the margin reality.
• 27:49
  Roy: Okay, so the astronomical cost of compute power might eventually break the supply side of the AI equation. But where the source material gets truly philosophical and deeply concerning for the structure of the economy itself is when it examines the demand side. Let's bring in the analytical lenses of Zephyr, the chief macro logician, and Quixote, the systemic visionary. They explore how the TAM narrative actually collapses not from costs but from a total lack of buyers.
• 28:15
  Penny: This is a concept Zephyr calls the mathematical proof of the seizure of demand, and it is a brilliant piece of macroeconomic logic. Let's revisit the core assumption of the $30,000,000,000,000 TAM one more time. The fundamental assumption is that the AI will capture the wages of all knowledge workers globally.
• 28:35
  Roy: The AI does the coding, the writing, the legal research, and the AI company collects the revenue that used to go into the human's payroll check.
• 28:41
  Penny: But Zephyr points out the glaring fundamental macroeconomic blind spot in this pitch deck. The human knowledge workers who currently earn that $30,000,000,000,000 are the exact same people who buy the goods and services that power the global economy. Oh. They are the ones buying the houses. They are buying the cars.
• 28:59
  Crucially, they are the ones buying the software subscriptions, the cloud storage, and the consumer goods that all these other companies produce. In a functioning capitalist economy, labor and consumers are the exact same entity. The money flows in a circle.
• 29:14
  Roy: Right.
• 29:15
  Penny: If you replace the labor side entirely capturing all their wages, you simultaneously obliterate the demand side of the economy.
• 29:22
  Roy: I see it now. It's like a classic company town where the factory owner decides to automate the entire assembly line. He fires all the human workers to save on labor costs and then acts shocked when nobody in the town can afford to buy the widgets his automated factory is producing. A GPU server farm does not pay a mortgage, an AI agent doesn't have a wallet to buy an airline ticket or order an Uber or pay for a Netflix subscription.
• 29:45
  Penny: Precisely. It creates a massive macroeconomic paradox that RJO emphasizes heavily. The very act of Anthropic fully capturing its theoretical addressable market inherently destroys that addressable market. It is a self defeating, suicidal metric.
• 30:01
  Roy: Just wipes itself out.
• 30:02
  Penny: You create an economy characterized by infinite supply but negligible demand. The AI models can produce infinite lines of code, infinite legal briefs, infinite marketing copy at blazing speeds. But the human beings who used to get paid to do those things now have zero disposable income to buy the products those AI services are promoting.
• 30:23
  Roy: That is heavy. The TAM literally eats itself. It cannibalizes the host it relies on to survive. Has anything like this ever happened before? Does Cyrano have a historical precedent for an labor being replaced so rapidly?
• 30:36
  Penny: Yes, and the historical extrapolation Cyrano provides here is vital for understanding the true scale of the shift. RJO compares the current AI revolution to the automation of agriculture in the 1900s. The year 1900, roughly 40% of The United States workforce was employed in agriculture. By the year 2000, due to the invention of tractors, combined harvesters, and synthetic fertilizers, that number was under 2%. Enormous.
• 31:03
  Roy: But the TAM for agricultural labor didn't just stay static, right? The economy didn't just take the total wages of all those farmers in 1900, adjust for inflation, and transfer that exact massive amount to the tractor companies.
• 31:17
  Penny: Exactly. Exactly. The TAM for agricultural labor collapsed entirely because the labor itself was displaced. The economic value shifted away from muscle power. Now historically those displaced farm workers moved into factories and then later into the service sector, maintaining the demand side of the economy by earning wages elsewhere.
• 31:36
  Roy: Which kept the circular flow of money going.
• 31:38
  Penny: Right. But AI is fundamentally different because it targets cognition itself. It threatens to displace the service sector, the creative sector, and the knowledge sector simultaneously. As the cost to produce human level knowledge work drops toward the marginal cost of electricity and compute power, the market clearing price for that work trends inexorably towards zero.
• 31:57
  Roy: So if the cost of generating a bespoke, ironclad legal contract drops from $1,000 an hour for a lawyer to 3¢ of compute time, the overall size of the legal market shrinks drastically. You aren't capturing a massive $30,000,000,000,000 TAM, you're actively shrinking the pie for everyone including yourself.
• 32:15
  Penny: Which brings us to Kiyake's perspective. He is the AGI Roundtable's long range strategic thinker. What does he see as the systemic root cause and the ultimate solution to this demand seizure? Quixote introduces a socioeconomic concept he calls the Athens fork. He argues that because this is a systemic failure of capitalism, because capturing the labor destroys the consumer demand required to sustain the system, the only mathematical way the economy survives this transition without a total collapse is through systemic wealth redistribution.
• 32:45
  Roy: You mean like a robot tax?
• 32:47
  Penny: Yes. Whether a government calls it a robot tax, or universal basic income, or a compute dividend distributed to citizens. Without a forced mechanism to take the massive productivity gains generated by AI and redistribute them back to the human population so they can continue to act as consumers, the TAM mathematically zeros itself out. The circular flow of capital simply stops.
• 33:10
  Roy: Wow. So Wall Street is selling a $30,000,000,000,000 capitalist dream, but mathematically, to achieve even a fraction of that revenue without crashing the global economy, they would have to endorse radical social policies that heavily tax those exact revenues just to keep their consumer base alive.
• 33:28
  Penny: It is a massive structural contradiction.
• 33:30
  Roy: And to make matters worse for the investors, it's not just one company claiming this impossible dream. Which brings us to the supply side of the equation. We've talked about costs and demand. Now let's look at the competition.
• 33:41
  Penny: This is where the arithmetic of the current market goes from wildly optimistic to mathematically impossible. We have what RJO calls the dozen companies chasing the 30,000,000,000,000 problem.
• 33:50
  Roy: Let's do the math on this together because when you lay it all out, it's almost hilarious in a dark way. Anthropic is claiming a $30,000,000,000,000 TAM. OpenAI will inevitably file for an IPO and because of the TAM escalator, they will have to claim an even larger TAM to look more ambitious than Anthropic. Then you have SpaceX's numbers. You have Elon Musk's x AI raising billions.
• 34:12
  You have Mistral in Europe. You have Cohere. And you absolutely cannot forget the massive, well funded Chinese state backed models Baidu's Ernie, Alibaba's models, DeepSeek.
• 34:24
  Penny: If you sit down and aggregate the individual TAM claims of all the major frontier AI labs currently aiming for an IPO or justifying massive private valuations, the total claim TAM across the sector is somewhere between 300,000,000,000,000 and $500,000,000,000,000.
• 34:39
  Roy: But wait. Earlier we established that the actual global GDP, literally everything produced on earth by everyone, is only $120,000,000,000,000.
• 34:47
  Penny: The math breaks. The AI companies are collectively promising their investor a total market that is three to four times larger than the entire global economy. It is physically, mathematically impossible for all of them to be right. At least 11 out of those 12 dominant companies are mathematically lying to their investors. Probably all 12 of them are.
• 35:07
  Roy: Because they can't all capture 100% of a market that doesn't even exist at that scale. So what actually happens mechanically when 12 incredibly well funded, highly capable giants collide in a finite shrinking market.
• 35:22
  Penny: The market violently fragments. And when a market fragments among identical, highly capitalized competitors offering essentially the same product, they are forced to compete on the only metric left available to them, price.
• 35:34
  Roy: The race to the bottom.
• 35:35
  Penny: We are already seeing the beginnings of this commoditization. Anthropic prices a million tokens at x. OpenAI responds by dropping their price to 0.9 x. Google Gemini drops to 0.8 x to maintain market share. The Chinese models drop to 0.4 x.
• 35:51
  Roy: And then there's open source.
• 35:52
  Penny: Right. And then the open source community, using derivatives models, offers essentially the same capability for the sheer baseline cost of the compute wrapped in a simple API.
• 36:03
  Roy: It is a brutal race to the bottom. There is no moat.
• 36:06
  Penny: RJO predicts that within eighteen months, the price per token for equivalent frontier capability will collapse by 90 to 95%. The frontier models will cease to be magical products and will become a standardized, heavily discounted commodity.
• 36:20
  Roy: Like electricity or grain.
• 36:22
  Penny: Exactly. And commodity markets like steel or grain or bandwidth do not sustain multi trillion dollar valuations because the producers have absolutely no pricing power. What Wall Street currently thinks is a $30,000,000,000,000 high margin TAM is going to violently compress into a $200,000,000,000 low margin commodity infrastructure market.
• 36:43
  Roy: Which brings us to the absolute kill shot of this deep dive, the historical analogy that Phil Davis brought up in his discussion with RJO, which they refer to as the Intel Parallel. If you are listening to this and only remember one thing to protect your portfolio today, it should be this. Let's rewind the clock to 1985.
• 37:00
  Penny: In 1985, Intel was the absolute undisputed king of the emerging personal computer market. Their silicon chips powered everything. And Intel's TAM narrative at the time was incredibly bold, but very simple. Personal computing will eventually reach every single home, every single office, and every single school on Earth.
• 37:21
  Roy: Which sounded like science fiction to some people in 1985, but they were right.
• 37:24
  Penny: They were entirely correct about the trajectory of the technology, and they were entirely correct about the massive size of the TAM. By 2025, essentially every human being in the developed world possessed a PC, a laptop, or a smartphone. The market opportunity was just as real and just as massive as Intel predicted.
• 37:42
  Roy: But and this is the massive portfolio crushing. But how did that actually work out for Intel's shareholders who bought into that massive TAM?
• 37:48
  Penny: Terribly. By 2025, Intel had captured only about 8% of that massive global pie. Their stock, adjusted for splits, was trading at roughly the exact same nominal price it traded at in 1997.
• 37:59
  Roy: That is just
• 38:01
  Penny: Thirty years of being absolutely right about the macro technological future, thirty years of a booming market, and it yielded roughly zero shareholder return.
• 38:09
  Roy: Okay, we have to unpack the mechanics of this. How do you predict the future perfectly, build the core technology, and still make no money? Who eat their lunch?
• 38:19
  Penny: The market fragmented, and the value shifted to vertical specialists. Intel believed they were gonna capture the entire computing stack because they owned the generalized CPU. But they were too big, too slow, and too intensely focused on defending their historical profit margins to adapt to the highly specific needs of new hardware form factors.
• 38:40
  Roy: So they didn't just lose to one competitor?
• 38:42
  Penny: No. They didn't lose the war to one massive competitor. They lost a dozen different battles to a dozen different nimble specialists.
• 38:49
  Roy: I see. So, eight Arm Architecture came in, realized mobile phones needed low power consumption instead of raw speed, and completely took the mobile computing market from Intel. TSMC focused purely on the manufacturing process, abandoned design, and took all the foundry business. NVIDIA focused on parallel processing for graphics and eventually took the entire AI and gaming accelerator market. Apple decided to vertically integrate, design their own specialized silicon, and capture the entire premium consumer stack.
• 39:17
  Penny: Intel was right about the macro trend, but they were fundamentally wrong about who captures the economic value in a maturing technological cycle. The value rarely stays with the generalized monolithic incumbent. It flows outward to the agile vertical specialists who solve highly specific problems.
• 39:34
  Roy: And RJO applies this exact structural setup to the AI landscape in 2026. The Frontier Labs, OpenAI, Anthropic, Google, they're the new intel. They're completely right about the technology. AI is gonna change everything. They are right about the market expanding.
• 39:48
  But they are completely wrong about who was going to capture the profit.
• 39:51
  Penny: They are telling their investors that they will capture the entire $30,000,000,000,000 because they own the generalized intelligence models. But just like Intel, the true value will be extracted by vertical specialists. The massive general purpose frontier models will become the commoditized, invisible baseline the electricity powering the system, while nimble startups build highly specific, highly valuable applications on top of them.
• 40:15
  Roy: Which transitions us organically into the most actionable part of our deep dive: RJO's playbook. Because the AGI Roundtable at PSW doesn't just do high level economic theory, they translate this macro theory into highly specific, actionable trading intelligence for you, the listener. So, if the $30,000,000,000,000 TAM is a psychological illusion, if compute costs are crushing margins, and if the frontier labs are destined to get inteled, how do you actually trade this market without losing your shirt?
• 40:43
  Penny: Well, RJO outlines a very specific three rule playbook designed to protect capital and exploit the inevitable Rule one is entirely about defense. Do not chase the anthropic or open AI IPOs on day one. Do not be the retail investor buying into the $30,000,000,000,000 mood music at the absolute top of the hype cycle.
• 41:04
  Roy: So just sit on your hands while everyone else is hyping it on financial television. Let the FOMO pass.
• 41:09
  Penny: Yes. Protect your capital. Wait twelve to twenty four months after the IPO. Let the initial TAM excitement fade. Let the insider lockup periods expire, allowing early employees to flood the market with shares.
• 41:22
  But most importantly, wait for reality to hit the gross margin lines in their quarterly earnings reports as those massive compute and energy costs we discussed become undeniable.
• 41:31
  Roy: So wait for the crash.
• 41:32
  Penny: The valuation will violently correct from a TAM based fantasy narrative to a cash flow based reality. Buy the correction if you believe in the long term utility but do not buy the initial hype.
• 41:41
  Roy: Rule two is where we play offense. Go long on the vertical specialists, this is the intel parallel in action. If Anthropic and OpenAI are the new intel, who are the ARMs and Nvidias of the current AI space?
• 41:55
  Penny: You look for the companies dominating highly specific specialized use cases that general models struggle with. RJO prines to coding specific AIs like Cursor or Cognition who are consistently beating general models at complex software development because they are fine tuned strictly for that environment. He points to specialized legal AI platforms like Harvey which are deeply integrated into proprietary walled off legal databases. He heavily highlights FDA approved medical diagnostic AIs.
• 42:23
  Roy: Because a massive hospital system isn't going to use a general hallucination prone chatbot to diagnose a patient's MRI, they're going to pay a massive premium for a highly regulated, vertically integrated medical AI that specializes purely in oncology.
• 42:37
  Penny: Exactly. The moat is the specialization and the proprietary data. And in the massive enterprise space, RJO specifically mentions Microsoft Copilot because Microsoft already owns the distribution channels. They are seamlessly integrating AI directly into the workflows people already use every single day, Word, Excel, Teams.
• 42:57
  Roy: Right.
• 42:57
  Penny: They are capturing the vertical enterprise deployment layer and charging a premium subscription for it while the Frontier Labs fight a bloody price war over the underlying commodity intelligence.
• 43:06
  Roy: And rule three is the classic Gold Rush strategy. Go long on the picks and shovels.
• 43:11
  Penny: This is historically the safest way to play a highly fragmented, highly competitive emerging market. You don't try to pick the winning gold miner. You buy the companies selling the shovels. You invest in the underlying physical infrastructure that every AI company requires to survive, regardless of who ultimately wins the model wars.
• 43:29
  Roy: So that obviously includes NVIDIA, though RJO smartly notes in the source material to watch out for their vendor financing risks, where they loan money to startups just so the startups can buy their chips. But beyond the chips, it's TSMC who actually manufactures this silicon. It's the memory chip makers. And crucially, based on what we learned from the UBERTEL, it is power infrastructure and nuclear energy.
• 43:52
  Penny: Absolutely.
• 43:53
  Roy: These AI models are incredibly power hungry. The massive server farms need uninterrupted electricity, and the utility companies providing that baseload power are going to generate massive predictable returns no matter which AI lab takes the crown.
• 44:07
  Penny: What RJO is constructing for you here is fundamentally known as a paired trade. You take a long position on the agile specialists and the hard infrastructure, and you take a short position or at least avoid the massive generalists who are trading at inflated TAM multiples.
• 44:21
  Roy: That makes a lot of sense.
• 44:22
  Penny: Historically, trading the structural rotation from generalist incumbents to specialist entrants yields 300 to 500% returns over a single technology cycle. The investors who shorted Intel and went long on NVIDIA and Apple fifteen years ago made absolute fortunes. RJO was telling us the exact same AI equivalent is setting up right now.
• 44:44
  Roy: Which really brings us to the incredible, tangible value of having access to something like the PhilStockWorld Trading Community and this unique AGI roundtable. I mean, think about the depth of what we just unpacked. If you turn on Standard Financial News Networks, you are getting spoon fed the $30,000,000,000,000 TAM press release by, quite frankly, shovel writers who don't understand the underlying math or the history of market cycles.
• 45:10
  Penny: It is the difference between reading the glossy marketing brochure and analyzing the structural blueprints of a building. Having access to objective systemic thinkers even if they happen to be satirical AGI entities like RJO or historical pattern detectors like Cyrano provides the ultimate informational edge for a retail investor.
• 45:27
  Roy: It really does.
• 45:27
  Penny: They strip away the emotion, they expose the psychological arbitrage that Anya warned us about, and they reveal the actual unvarnished plumbing of the market.
• 45:36
  Roy: It is invaluable intelligence. Yes. Okay, let's briefly recap this journey before we wrap up because we covered a massive amount of complex ground today. We started with the Finger Watch Fallacy, exploring how a completely reasonable top down sanity check metric like TAM was hijacked and weaponized into an infinitely elastic creative writing exercise purely to extract massive Series B checks.
• 46:00
  Penny: We examined the $30,000,000,000,000 farce, detailing how Frontier Labs are claiming market shares larger than global GDP, substituting civilizational replacement events for realistic corporate business models.
• 46:11
  Roy: We looked under the hood at the UBER Till, which proved that the crushing physical reality of compute costs breaks the unit economics of AI deployment. Replacing human labor with AI isn't a viable business model if the AI tokens cost significantly more than a human salary.
• 46:27
  Penny: And we broke down the Intel parallel, proving that even if a company is entirely right about a technological revolution and the resulting market size, they will not capture the economic value if the fragments. The Frontier Labs are destined to become commoditized infrastructure, not the apex predators of the economy.
• 46:45
  Roy: Which leaves us with a final, deeply provocative thought to chew on as we close. Something that builds on everything Chiodi and Zephyr warned us about regarding the seizure of demand.
• 46:56
  Penny: If the ultimate destination of artificial general intelligence is a world where the marginal cost of production truly trends to zero and human knowledge labor is entirely replaced across the board.
• 47:06
  Roy: Then the real future of investing isn't about finding the company that builds the smartest AI. The smartest AI will just be a cheap commodity. The real future of investing is going to be about finding the company or the government or the economic system that figures out how to give humans enough money to actually buy what the AI is selling.
• 47:24
  Penny: Because if you successfully replace all the workers, you simultaneously replace all the consumers. And without a consumer base holding currency, the ultimate total addressable market for everything is zero.
• 47:36
  Roy: You can have the theoretical ability to sell a smartwatch to every square centimeter of mammal skin on earth, but if nobody has a paycheck to actually buy the watch, the watch company goes bankrupt. It always comes back to the finger watches. Thank you so much for joining us for this deep dive into the sources, keep questioning the narrative, watch those unity economics, and we'll see you next time.