The revenue number Anthropic showed prospective investors this week was large and the growth rate attached to it was smaller than the one before, and that combination has some of them recalculating what the company is worth.
People familiar with the matter said investors who attended the company’s Wednesday investor day discussed a valuation of about 1.5 trillion dollars, down from roughly 2 trillion dollars that had circulated earlier. Existing holders and prospective buyers of shares in a listing were among those weighing the lower figure.
Anthropic told the group that revenue in the third quarter rose 45 percent from the previous quarter to 16.8 billion dollars, and that the quarter was profitable on an adjusted basis. In the second quarter, revenue more than doubled from the first, reaching 11.6 billion dollars.
The deceleration is the part investors focused on. A company growing at triple-digit rates quarter over quarter is priced as though the rate will persist for years, and a slide to a high double-digit rate changes the arithmetic of how much cash the business will produce before growth settles. Nothing about the third-quarter figures suggests a contraction; revenue still rose by more than five billion dollars in three months, a figure most software companies never reach in a year.
The cost side of the business complicates the read. Developing and serving frontier models requires computing capacity bought years in advance, and the accounting for those commitments spreads over the life of the hardware. Adjusted profitability of the kind the company reported on Wednesday excludes the items that make the underlying economics harder to interpret, which is why investors asked about cash flow rather than earnings.
Timing is the other live question. The company has been reported to be preparing to begin its listing roadshow as early as the week of November 9, with trading potentially starting before Thanksgiving, and a listing by the end of the year at the latest. An offering calendar that tight means the price has to be settled quickly, and the buyers who set it are the same investors sitting in the room this week.
The market backdrop will not help. Semiconductor shares fell sharply during the week, hedge fund portfolios recorded one of their worst days since the pandemic, and credit investors have begun to price the risk of the companies that build data centers. A listing of a company whose economics depend on continuing to buy computing capacity into a wobbling trade in computing capacity is harder to place than the same listing would have been a month ago.
Competitive pressure adds another variable. Anthropic sells access to models that compete with offerings from a handful of rivals, and buyers of enterprise software rarely commit exclusively to one vendor. Pricing on the largest contracts is negotiated, and the company must keep spending to stay near the capability frontier, which is a cost that does not fall when growth slows.
What the investors are being asked to underwrite is a business with unusually high gross margins and unusually high fixed costs, in a market where the leading position changes hands every few quarters. That profile has supported large valuations when revenue is accelerating, and it invites a discount when the second derivative turns negative.
The comparison the market reaches for is not another software company but the capital-intensive businesses that supply it. Data center operators, equipment makers and memory producers all trade on expectations of a buildout cycle, and their multiples compress when the cycle is questioned. A model developer that buys from all of them inherits some of the same sensitivity.
Some of the investors said the third-quarter figures eased their concerns about the scale of the spending required to stay competitive, even as the growth rate disappointed them. That is a reasonable position: the company showed it can convert spending into revenue and profit, which is the question that matters most for a business at this stage.
What the offering will not resolve is how durable the margins are. Serving costs fall as models become more efficient and rise as the models customers ask for become more capable, and no developer has published a long enough record to show which effect wins. Buyers of the listing will be making that judgment with three quarters of data and a shrinking runway before the shares start trading.
Early investors in the company have a separate set of incentives. Venture funds that bought at earlier valuations have marks tied to the last round, and a lower price at listing takes an unrealized gain and makes it smaller. Funds raising new capital from their own investors have an interest in a clean exit, which supports taking a price that can be defended after the first day of trading rather than one that relies on continued momentum in a falling market for related equities.
The employees who hold equity in the company have the longest horizon of anyone in the room and the least ability to reduce their exposure before a listing. Their compensation depends on a price set in the next few weeks, which is a further reason the figure under discussion this week matters beyond the investors who negotiated it.
