Key Highlights

  • Anthropic's filing sets a USD 70 to USD 78 price range for 825 million shares, with pricing reported for mid-November.
  • Anthropic's prospectus reports first-half Revenue of USD 16.3 billion, nearly 14 times the figure a year earlier.
  • Debts and leases at the five largest Data Center owners equal 30% of book value, or 6% of market value.
  • Direct US exposure to AI spending is put at roughly 0.3% of GDP, because most equipment is imported.

A Listing Arrives as the Financing Question Sharpens

Anthropic has set terms for its Nasdaq listing. Its filing lists a price range of USD 70 to USD 78 a share for 825 million shares, and reports point to the week of Nov. 16 for pricing. Coverage of the filing puts the implied valuation at roughly USD 1.3 trillion, and up to USD 1.4 trillion at the top of the range.

The timing is awkward. AI models keep improving faster than almost anyone forecast, yet the market that finances them has turned jumpy, because investors are asking whether AI revenue can sustain spending on this scale. The listing puts that question in front of every investor at once.

How Large the Bet Has Become

The scale of the wager explains the nerves. Anthropic's prospectus reports revenue of USD 16.3 billion in the first half of 2026, compared with USD 1.2 billion a year earlier, and a run-rate of about USD 81 billion at the end of October. The company defines run-rate as usage fees over the trailing 28 days multiplied by 13, plus monthly subscription fees multiplied by 12, and the filing says it is not a forecast and that the growth rate is anticipated to decline over time. Public investors are being asked to value that growth.

Anthropic and OpenAI sit at the top of the Supply chain. Their sales feed data center builders, chip designers and chipmakers, so a shortfall in revenue or pricing would travel down the chain. OpenAI has been reported to be leaning toward waiting until next year for its own listing.

What Past Technology Booms Show

Canals, railways and the internet all produced euphoria followed by despair, even though the technologies proved transformative. Investors who financed fiber-optic cable in the 1990s lost heavily, while companies that built businesses on top of that cable prospered. Backers of new technologies have often misjudged who captures the profit.

The same question applies now: who captures the profit from AI. The prospectus gives investors a view of Anthropic's finances, including compute and infrastructure expense that rose to 58% of revenue in the third quarter from 48% in the second.

The gloomiest comparison is with the railway bust of 1873, the last Investment binge of comparable size relative to GDP. Disappointing returns then set off a financial calamity, Deflation and the longest American Recession on record.

Why the 1873 Comparison Is Disputed

Several features differ. Most American Capital spending on AI goes to imported equipment such as chips, which puts direct exposure of the US economy at perhaps 0.3% of GDP. Taiwan, whose economy grew 14.4% in the third quarter from a year earlier, is more exposed. In the 1870s there was no central bank, and policy tightened after the crash.

Financing also differs. The worst crashes, 1873's included, followed booms funded mostly by debt, with excessive household Credit the strongest warning sign. AI companies are borrowing far more than two or three years ago, when tech giants' cash flows paid for most investment. Even so, the accumulated AI Capital Stock is still mostly financed with equity. The five largest data center owners carry debts and Lease commitments of around 30% of book value, or 6% of market value, and their older businesses remain highly profitable. Nvidia's Leverage is negligible. AI-related Debt has not pushed lending to the US nonfinancial sector above its long-run trend relative to GDP.

Where Losses Would Land

In a crash, households with Equity losses would spend less. Highly leveraged companies, including many data center startups that rent out computing power, could fail. One assessment describes a shallow American recession comparable to 2001. A deeper downturn would require another shock, such as a bond-market panic or a war.

Politics adds another variable. Of 41 local data center ballot measures in 10 states on Nov. 3, the results skewed almost unanimously against development, according to a review by a law firm, so the pace of construction depends on voters as well as on financing.

What the Debut Puts to the Test

The offering gives public markets their first direct read on how the businesses at the top of the AI supply chain are valued, and several questions remain open. The financing mix is the clearest. So far the buildout rests mostly on equity, and the data center startups that borrow against chip-backed leases are the most leveraged part of the system. Revenue is the second. Revenue has to keep growing for the money spent on chips and data centers to earn a return, and Anthropic's first public financial reports put that growth rate in view every quarter. Policy is the third, because federal proposals aimed at data centers or AI developers, and local restrictions, change the pace of construction regardless of financial conditions. Nvidia's quarterly results on Nov. 17 offer the clearest public read on whether orders from data center builders have shifted, and the Offer Price and first weeks of trading in Anthropic show how much investors are willing to pay for the top of the chain.