Anthropic priced itself at a $2 trillion valuation based on ten months of revenue, which could be the most absurd principle in the investment world.

Author: Byron Gilliam (Blockworks)

Translated by: TechFlow

TechFlow Highlights:In its highly anticipated S-1 filing, frontier AI giant Anthropic has orchestrated a breathtaking feat of valuation wizardry: multiplying the last 28 days of API consumption fees by 13, adding the current monthly subscription fees multiplied by 12, and even leaving a glaring blank space at the core, waiting for the latest single-month data released in October to back an astronomical potential valuation of $2 trillion.

Byron Gilliam, Senior Macro Analyst at Blockworks, draws on the Heisenberg Uncertainty Principle from quantum mechanics to sharply expose the financial markets’ most ironic valuation paradox: the more precisely a financial metric is calculated, the less useful it becomes for measuring a company’s long-term intrinsic value. Unable to predict cash flows over the next few decades, investors settle for anchoring to the next year’s price-to-earnings ratio; when faced with massive losses, they concede to the price-to-sales ratio; and in the frenzied AI capital arms race, the market has even regressed to relying solely on annualized data derived from a single month.

The stock market euphoria for global semiconductor giants now hangs on a single blank line in a prospectus. This marks the most absurd valuation exercise in investment history, yet it stands as the resigned compromise of capital with no other options during a manic cycle.

The Heisenberg Principle of Investing

According to the laws of quantum mechanics, matter exhibits wave-particle duality: every discrete particle also exists as a probability wave.

This makes particles elusive. To determine a particle’s exact position, you must collapse its wave nature into a single point, but this causes it to lose a definite wavelength; and to know its momentum, you must allow the wave to spread freely through space, which causes the particle to lose a fixed position.

This is the famous Heisenberg Uncertainty Principle: the more precisely you know a particle’s position, the less you know about its momentum.

A similar phenomenon plays out in financial markets.

The foundational principle of investing is straightforward: the intrinsic value of any asset equals the net present value of all its future cash flows discounted to the present.

By logical deduction, the most accurate method to value a company’s stock should be forecasting all its future cash flows.

Yet this is a nearly impossible task. Who can accurately predict how much profit a company will generate ten or twenty years down the line? Who can even assert whether it will still exist in that world? Typically, the farthest we can stretch our forecasting capabilities is to project earnings for just one or two years ahead.

Consequently, our trading logic in secondary markets follows this constraint. Stocks are typically valued by anchoring them to a multiple of expected earnings for the coming year (the price-to-earnings ratio, or P/E), since it is the only metric that can be calculated with reasonable precision.

However, precision comes at the steep cost of utility: a short-term earnings forecast for just one year captures only a minuscule fragment of a company’s prolonged intrinsic value.

This is the trade-off investors must accept: the more precisely a financial metric is measured, the less practically useful it becomes in determining an asset’s true intrinsic value.

We might as well call it the “Heisenberg Principle of Investing.”

Take global memory chip leader SK Hynix, for example. It trades at roughly a 4x price-to-earnings multiple relative to expected earnings over the next year, which superficially looks inexplicably cheap.

But how useful is this metric as a gauge of the stock’s intrinsic value? Actually, very little. In the semiconductor industry, defined by sharp swings and intense cyclical volatility, massive profits in one year can quickly devolve into bottomless losses the next.

Yet it is precisely this extreme cyclical volatility that forces us to rely on short-term forecasts to value SK Hynix: the true core driver that determines the stock’s long-term intrinsic value is the long-cycle supply-demand equilibrium in global memory chips—a dynamic that simply cannot be accurately predicted in reality.

And when a company isn’t even generating positive profits today, the difficulty of valuation rises exponentially.

Faced with persistently loss-making companies, the usual approach is to backtrack up the income statement (P&L) until you find a positive figure to apply a valuation multiple to, such as revenue.

For instance, cybersecurity software giant CrowdStrike is essentially operating without meaningful profits. Nevertheless, it still commands a massive market cap of $274 billion because investors choose to value it at roughly a 40x price-to-sales (P/S) multiple based on expected revenue for the next year.

A 40x P/S multiple is certainly expensive, but cybersecurity is a blazing hot sector, and in the absence of profitability, revenue remains our only viable anchor for valuation.

But what if even next year’s revenue becomes too elusive to forecast? In such an extreme scenario, we might have to fall back on current actual financial data.

Consider frontier AI leader Anthropic, which is currently heavily steering investors toward valuing its equity based on a multiple of its “current revenue.” Yet the precision of this metric is far murkier than it sounds. What exactly does “current” mean here? Revenue for the full year? Revenue for the month? Or revenue for just today?

In the highly anticipated S-1 filing, Anthropic provided its own custom formula for calculating “run-rate revenue”:

This is an operational metric intended to reflect our estimated annualized value from revenue-generating activities as of the measurement reference date. It is calculated as follows: (1) API consumption fees over the consecutive 28-day period preceding the reference date multiplied by 13; plus (2) total monthly subscription fees as of the reference date multiplied by 12, summed together.

In other words, Anthropic wants investors to annualize its API revenue from the most recent month (a pay-as-you-go model) and also annualize its most recent day of subscription fee revenue (a periodic renewal model).

Then, it expects you to multiply this amplified, annualized number by an ambitious multiple—likely 15x or even 20x.

But hold on—all of this won’t be revealed until November!

In yesterday’s disclosed prospectus, notably left blank was a specific placeholder. Anthropic promises to immediately fill in this figure once the annualized data for October is compiled:

As often joked in the AI circle, this small blank area is “bearing an unbearable weight.” The success or failure of Anthropic’s century-plus IPO hinges almost entirely on the number that will eventually emerge from this blank space.

This is also the core logic behind why global semiconductor stocks rallied collectively this week: market participants firmly believe that if Anthropic didn’t have absolute confidence that October’s data would be impressive, it would never have boldly carved out this specific blank space in the prospectus.

If the logic holds, Anthropic will raise significantly more capital to buy semiconductor chips in bulk; and as long as it purchases more chips, manufacturers like SK Hynix can secure healthier profits, thereby lowering their projected P/E multiples—the very mechanism we initially used to value these semiconductor giants.

Is this really a healthy way to invest? Can the operating results of a single month from another private company truly leverage and revalue the entire sector’s stock market capitalization?

The answer is clearly no. Crude annualization of just one month’s revenue has completely slipped to the extreme opposite end of the “Investing Heisenberg Principle”—where a metric achieves perfect precision but zero practical utility.

Using run-rate revenue to price a behemoth approaching $2 trillion can only be described as the most absurd and outrageous valuation gimmick.

But unfortunately, it remains the best—and only—tool we have at our disposal right now.