A $2 trillion valuation for a private artificial intelligence company sounds designed to end conversations rather than start them. On CNBC this afternoon, Lo Toney, founding managing partner at Plexo Capital, argued the opposite.
If you accept a certain revenue run rate at listing and apply a multiple borrowed from the public AI cohort, the figure lands where the framework says it should.
Toney disclosed on air that his firm holds Anthropic from the Series D that closed in January of 2024 at an $18 billion valuation, so he is speaking about a position he owns.
That disclosure does not disqualify the argument, but the argument has to survive on its own terms, because a valuation multiple is only as defensible as the revenue it multiplies and the durability of the margins behind that revenue.
Anthropic is not listed; its shares are not available to ordinary investors today, and the October timing Toney referenced is an expectation described in the segment rather than a scheduled event.
The Framework Toney Laid Out
The mechanics are simple. Toney said that “when you start to apply these multiples to Anthropic… you can come down to a multiple anywhere, you know, call it about 25x. And if in October, when everyone expects the IPO to hit, they’re at about 80 to 85 billion in the run rate for revenue, then all of a sudden, you know, 2 trillion starts to look fairly reasonable.”
He anchored the growth story by saying Anthropic has “truly bucked the law of large numbers” in its revenue trajectory over the past three years. Whether that pace holds through the summer is the assumption on which everything else rests.
Two of Anthropic’s largest backers have already told the public market that the private valuation has been repricing sharply. Microsoft (NASDAQ:MSFT | MSFT Price Prediction) recognized a $3.2 billion gain from its investment in Anthropic in its fiscal fourth quarter, and Alphabet (NASDAQ:GOOG, NASDAQ:GOOGL) reported a $99.03 billion gain on equity securities in the second quarter, with the Anthropic stake widely understood as a meaningful contributor. Those marks reflect what strategic partners will pay for optionality, not what a public float has to clear.
Why the Public Comparables Do Not Cleanly Translate
The 25x figure only holds if the comparables behave like Anthropic. NVIDIA (NASDAQ:NVDA) trades at a price-to-sales ratio of about 22x, with non-GAAP gross margins of 75% and operating margins of around 65.6%, on data center revenue that grew 92% year over year in the most recent quarter. Microsoft trades at a price-to-sales ratio of about 11x with a 45.1% operating margin. Alphabet trades at a price-to-sales ratio of about 9x with a 34% operating margin. A 25x multiple sits above all three, and each is profitable at scale.
Anthropic operates without that profitability cushion. Reporting circulating the same day Toney appeared observed that Anthropic has barely turned a profit, which is the necessary floor for any comparison. A multiple is a claim about the future cash flows the market will pay for today, and multiples above those of profitable incumbents require faith that growth lasts long enough to close the profitability gap. That faith is easier to sustain in a private round with strategic partners than in a public market with generalist money.
The Margin Question Is the One That Matters
The part of the segment that warrants the most attention is the caveat Toney himself supplied. “As you move up into the models where Anthropic and OpenAI’s core business is, we do believe that that is a place where there will be some pressure on the margins. You can see some of that coming from the open models, and the orchestration companies are now doing to route the right task to the right model.”
Enterprise buyers increasingly treat frontier models as substitutable. Satya Nadella described exactly this architecture in July, saying “any given model at any given time is swappable” across a catalog that now spans over 11,000 models.
Toney’s math is the right way to think about a private company at the moment of listing, because a run rate applied to a comparable multiple is what public markets do every day. What the framework understates is the question of durability.
A 25x sales multiple is a bet that terminal margins will compound toward something like the public leaders, and if orchestration and open weights compress the pricing frontier that model providers can hold, the terminal margins will not resemble those behind Nvidia or Microsoft today.
If Anthropic prices at $2 trillion in October, the interesting question for ordinary investors will be whether the revenue that justified it still carries the same margin two years later.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks