Aswath Damodaran, the New York University finance professor known as the dean of valuation, has put a price tag of $1.475 trillion on Anthropic, placing his estimate just below the reported $1.5 trillion to $2 trillion range for the artificial intelligence company’s planned initial public offering. His detailed valuation, published after Anthropic filed its prospectus, offers a rare independent check on one of the largest listings ever attempted.
A valuation built on ambitious growth
Damodaran’s central case assumes Anthropic’s revenues approach a trillion dollars by 2036, a trajectory requiring sustained hypergrowth for a decade. He applies a target operating margin of 27.5 percent, after a five-point regulatory drag on an underlying margin of 32.5 percent. The model also assumes reinvestment of about one dollar for every three dollars of additional revenue.
His cost of capital of 9.74 percent reflects the risk of a loss-making company and the scale it must reach to justify its price. On those inputs, equity value lands at $1.475 trillion, below the bottom of the offer range. Investors weighing the offering and paying the top of that range would need stronger growth, wider margins, or both.
The range is strikingly wide
The most striking feature is not the headline number but the spread around it. In the worst case, equity could be worthless, with a negative value of $373 billion, if 2036 revenues reach only $200 billion with margins of just 10 percent. That scenario shows how quickly a capital-hungry AI business can destroy value if growth stalls.
At the other extreme, revenues of $1.5 trillion in 2036 with margins of 40 percent would support a value of $4.04 trillion, more than double the offer range. The distance between worthless and $4 trillion shows Anthropic is not priced like a mature company with predictable cash flows. It is priced on a bet about the AI industry’s future.
What the prospectus reveals
Damodaran draws heavily on Anthropic’s prospectus, filed on 26 October 2026 and running to 416 pages. Revenues in the first half of 2026 were about 14 times the level of the same period in 2025. Yet growth slowed in the third quarter of 2026, as OpenAI gained share, computing capacity constrained supply, and some enterprises shifted to cheaper open-source models.
Enterprise adoption shows the same tension between momentum and concentration. Enterprises spending more than $1 million a year on Claude rose to 1,900, up from fewer than 100 a year earlier. At the same time, the top five customers account for 25 percent of revenue, leaving Anthropic exposed if a large customer changes strategy.
The cost picture is improving but remains daunting. The adjusted operating loss exceeded $8 billion in the first half, while the reported loss reached $80 billion after a $72 billion non-cash charge. The company claimed a profit last quarter, and compute costs fell from over 250 percent of revenues to 48 percent quarterly, helped by large chip deals.
Anthropic’s balance sheet looks formidable by contrast. Its cash balance stood at $114 billion, likely to be bulked up by about $100 billion from the offering itself. That war chest sits alongside compute commitments of more than $600 billion, a BB+ bond rating, and a 10 percent estimated probability of failure, with half of value lost in that event.
Governance draws sharp criticism
Damodaran is sharply critical of Anthropic’s governance, which he argues weakens shareholder control in ways that should trouble public investors. The company has five share classes with different voting rights, and its founders control 50.1 percent of the vote no matter how many additional shares are issued. A benefit trust, rather than ordinary shareholders, picks the board.
As a public benefit corporation, Anthropic can balance shareholder interests against other interests even when doing so costs performance, Damodaran notes. For investors, that means buying economic exposure without the usual levers to hold management to account. The structure may protect the mission, but it removes a traditional safeguard for minority shareholders.
The valuation should therefore be read as one expert’s reasoned judgment, not a market price or prediction. Damodaran himself cautions that he may be massively wrong, and the breadth of his own range supports that humility. Investors weighing the offering must judge his growth and margin assumptions against the governance discount his critique implies.
Why It Matters
The Anthropic listing tests how far public markets will go in funding the AI buildout, and a valuation from a figure of Damodaran’s standing gives that debate a useful anchor. If his central case is broadly right, the reported pricing range is ambitious but not absurd. If growth slows further or margins compress, the downside in his model is severe.
For the wider AI sector, the exercise clarifies what must be true for today’s valuations to work: revenues on a scale few companies have reached, margins closer to software than infrastructure, and compute costs that keep falling as a share of sales. Anthropic’s prospectus shows progress on costs, but the third-quarter slowdown shows how fragile the path remains.
Background
Anthropic, maker of the Claude family of models, has moved rapidly from research laboratory to one of the most valuable private companies in the world, and its listing is closely watched across technology and investment, including who stands to benefit most from it.
Source: Damodaran’s valuation post
0 Comments