Anthropic Jumps the IPO Shark with Garish TAM Estimate

When are we going to call time on the bullshit and blatant dishonesty of the putative leaders of AI behemoths? It’s at least partly a rhetorical question, for it assumes that time needs to be called on the worst offenders. What’s at question is when we make the call. I will argue here that the time in question ought to come sooner rather than later.

I nearly spat out my coffee yesterday when reading that Anthropic, girding for what it hopes will be an unprecedentedly lucrative IPO, somehow estimates that its total revenue opportunity exceeds $30 trillion. No, that is not a typographical error. I did not, as a result of a sudden involuntary spasm or the effects of tendinitis in my right thumb, type a “t” rather than a “b.” We’re talking trillions here. Carl Sagan, who spoke and wrote of “billions and billions” in a decidedly different context, would be astounded.

Invoking technical terminology and the vernacular of the people of Britain, I submit that Anthropic is talking bollocks. How did the carnival barkers preparing this IPO extract such astronomical numbers from their backsides? As you will see, and as I’m sure you’ve already guessed, the formulation of Anthropic self-referential revenue potential was a result more of performance art than mathematical rigor.

Abused and Tortured TAMs

According to a report in the Wall Street Journal, Anthropic says the figure represents an estimate of its total addressable market (TAM), assuming that the company captures the entirety of the available market opportunity. In other words, Anthropic asserts that the TAM represents a theoretical scenario in which it wins 100% market share. I realize that companies use and abuse TAMs this way all the time, but that doesn’t mark the malpractice right, reasonable, or particularly useful. All it means is that TAM malpractice is rife.

Do you know of any market worth owning, in a true market-based economy (which the U.S. might no longe possess under the current, egregiously corrupt, coin-operated administration), in which any single vendor claims the whole market, eliminating any and all competition? It’s entirely fictitious, something that will never and should never happen.

So, the first objection to this pre-IPO posturing is that no vendor ever gets 100% of any market subject to competition. In practical terms, rather than as part of a marketing pitch, a TAM should only be used as a representation of an entire market opportunity, available to all competitors (plural) in a given market. Anthropic is not going to get 100% of the AI market. It doesn’t have 100% of the market today, and there is no realistic scenario in which it will claim 100% of the market in the future, no matter how it unfolds.

Yes, Anthropic is technically correct in calling what it has adduced a TAM, but that doesn’t make these numbers useful as an approximate yardstick for gauging the company’s real-world revenue potential. For that, Anthropic should offer something more practical, such as a serviceable addressable market (SAM) — the portion of a TAM that its offerings can serve today — combined with Serviceable Obtainable Market (SOM), the portion of the SAM that the company realistically expects to claim within, let’s say, three to five years. (Anthropic could stretch that a little further to seven years, if it must, but generally forecasts become less accurate, and therefore less useful, the further out they extend.) In aggregate, the triumvirate of the TAM, SAM, and SOM can be reasonably credible tools for companies to use when pursuing financing and preparing for IPOs.

An unrealistic TAM, wielded indiscriminately, does everybody a disservice. When you see a TAM such as the one Anthropic has proffered, your only response should be incredulity, perhaps compounded by indignation. Your intelligence deserves respect, even in the AI era.

Anatomy of a Sales Pitch

Let’s look at how Anthropic put together this TAM. The Wall Street Journal informs us that sources told it that, to quantify its TAM, Anthropic looked at “the full scope of work that could be completed with AI models.” From the article, we do not know how Anthropic defined the full AI scope of work, nor how generously Anthropic viewed the current and future capabilities of its AI portfolio. You can be reasonably confident that the company took considerable liberties on both fronts.

This excerpt from the Wall Street Journal article is telling:

Anthropic more than doubled its revenue to $11.6 billion in the second quarter. To put its more than $30 trillion vision in context, the 191 technology companies in the S&P 1500 brought in $2.4 trillion in revenue last year, according to FactSet.

Drinking All the Milkshakes

We don’t even know whether the current revenue number is entirely accurate. We have to take on trust from a company that is now showing us that it can’t be trusted.

At the risk of repeating myself, this is bullshit or madness, perhaps both. Are we supposed to believe that Anthropic will effectively rule hegemonically over the entire economy, conquering nearly all other industries and vocations? Or are we supposed to believe that AI will produce a ludicrously high economic productivity exponent, despite drinking the milkshakes of most industries in its thrall. Where would all these trillions come from? Would there be growth for anybody else?

The TAM Anthropic has fabricated is neither a well-constructed top-down model nor a rigorous bottom-up model. What is it then? We must conclude that it is merely a storybook exercise in IPO marketing. Anthropic, and presumably its investors and investment bankers, wanted an obscenely high TAM number, even bigger than the one that Elon Musk and his confrères concocted as a launchpad for the SpaceX IPO. This is pure one-upmanship, a my-TAM-is-bigger-than-yours boardroom boast.

Musk, in his characteristically modest and reserved manner, declared that the SpaceX TAM, pegged at $26.5 trillion, was “the largest actionable” market in “human history.” The fact, it was largely made-up, including projected AI revenues that far exceed SpaceX’s current standing in the AI market and any rational estimate of how much of that market it will capture in years to come.

Anthropic believed it had to go above and beyond the level of pre-IPO TAM hyperbole that SpaceX had established. Given that the bar was purely imaginary, nothing in our material world prevented Anthropic from clearing it.

We Can and Must Do Better

This is where we find ourselves in the third decade of the 21st century. We lie to each other all the time. In the process, how much damage are we doing, not only to others but to ourselves?

Don’t be deceived by these sideshow carnival barkers, folks. We should use history as an approximate guide, recognizing that historical patterns never repeat identically, though they share certain behavioral traits. They share those behavioral traits because markets, which are abstracted representations of commercial transactions, issue from human agency.

As the Wall Street Journal article recounts, Uber, when it went public in 2019, estimated its market opportunity at $6 trillion. WeWork, of all companies, cited a TAM of $3 trillion before abandoning its IPO.

WeWork, a former high-flying unicorn whose IPO crashed and burned in a pyre of delusions, demonstrates the intoxicating allure of imagined riches. In a New Yorker article, published after the implosion of the IPO in late 2019, a former employee said the following:

“In retrospect, there’s no way this could have worked,” one employee, a software engineer, told me, sounding weary. He brought up the marketing expert Scott Galloway, who has compared cheap capital to a drug. “People were high. There’s not a human being in America who doesn’t look at the number forty-seven billion dollars”—WeWork’s valuation in January—“and not get goosebumps. It seems insane now, but at the time it made so much sense.”

Learning the Hard Way

Will we utter the same words about the AI market bubble when we look back retrospectively a few years from now? Are we succumbing to the same insanity again? Sure, AI has a more substantive value proposition than WeWork had, but it seems, in the current feral moment, just as susceptible to delusions of grandeur.

Since we’ve mentioned the TAM that preceded Uber’s IPO, perhaps we should see how closely that company has tracked its overly ambitious forecast. Earlier this month, Uber announced its second-quarter financial results for 2026. Quarterly revenue grew 12% YoY to reach $14.2 billion. (Growth was lower, 11%, on a constant-currency basis.) There is no way, ever, that Uber will attain its pre-IPO TAM number of $6 billion. I’m sure nobody at Uber today speaks of that number.

I hear your objection. You say AI is different. That’s what we always say. In certain respects, perhaps AI is different, but the people pushing these IPOs are cut from the same cloth as those who promoted Uber’s IPO and the aborted WeWork IPO. They’re selling maniacally, without the healthy restraints of inhibition or scruples.

It isn’t technology that produces market bubbles. It’s the people behind them. The people behind the IPOs of the AI behemoths appear to have learned nothing from the experiences of IPOs past. They’ll have to learn the hard way. As for the rest of us, caveat emptor.