Quarterly profit underpins hopes for valuation of up to USD 2 trillion
Slowing revenue growth and contractual payments limit investment capacity
Early shareholder selling could intensify downward pressure on shares

U.S. artificial intelligence (AI) company Anthropic has begun the listing process by publicly filing its registration statement for an initial public offering (IPO). Buoyed by surging revenue and a quarterly operating profit, it is seeking a valuation of up to USD 2 trillion (approximately USD 2.04e+3 billion). But a recent slowdown in revenue growth is raising questions about the momentum needed to sustain that lofty valuation. Its capacity to invest in accelerating growth is also limited. Huge payment obligations under existing infrastructure contracts mean a substantial share of the IPO proceeds is likely to go toward fulfilling prior commitments. With existing shareholders also set to sell shares early in the listing, the stock could face selling pressure before the growth expectations embedded in the IPO price translate into results.

Anthropic Begins Nasdaq Listing Process

According to The Wall Street Journal (WSJ) on the 26th (all dates hereafter are local time), Anthropic publicly filed its S-1 registration statement for an IPO with the U.S. Securities and Exchange Commission (SEC) that day. The filing showed first-half revenue of USD 16.3 billion (approximately USD 16.6 billion), roughly 14 times the year-earlier figure, while its operating loss narrowed from USD 2.1 billion (approximately USD 2.14 billion) to USD 693 million (approximately USD 707 million). In particular, the company turned profitable in the second quarter, posting operating income of USD 974 million (approximately USD 993 million). Second-quarter revenue rose about 144% from the previous quarter, while computing and infrastructure costs increased about 65% over the same period. In other words, as service sales grew, computing and infrastructure costs accounted for a smaller share of revenue.

The first-half net loss was USD 80.7 billion (approximately USD 82.3 billion), consisting mostly of accounting charges arising from changes in the value of convertible financial instruments. Of that amount, USD 72.4 billion (approximately USD 73.8 billion) was a noncash loss from fair-value remeasurements of convertible bonds and similar instruments. As the company's valuation rose, so did the assessed value of financial instruments that investors would later convert into shares, with the increase recognized as an accounting expense. Anthropic is pursuing a Nasdaq listing, but did not disclose the offering price range or number of shares to be issued in this filing. Market estimates put its expected valuation at USD 1.5 trillion–USD 2 trillion (approximately USD 1,529.65 billion–USD 2.04e+3 billion).

A valuation of up to USD 2 trillion already reflects considerable expectations of continued rapid growth. Third-quarter revenue of USD 16.8 billion (approximately USD 17.1 billion), disclosed at an investor briefing on the 14th of this month, translates into an annualized figure of USD 67.2 billion (approximately USD 68.5 billion). The upper end of the expected valuation is 30 times that amount. Revenue used as the basis for the valuation also includes fees paid to cloud distributors. Anthropic records the full amount customers pay through cloud platforms as revenue and treats the platforms' share as an expense. About 48% of its first-half revenue this year came through cloud marketplaces, meaning fees paid to distributors were included in revenue.

First-Half Operating Loss Despite Quarterly Turnaround

Maintaining a high valuation, however, requires steady growth in revenue and profit, and revenue growth has recently been slowing. Anthropic's third-quarter revenue rose about 45% from the previous quarter, roughly one-third of the second-quarter growth rate. Customer spending has also been growing more slowly than service usage. Data from corporate spending management company Ramp showed that Anthropic's token usage increased nearly 90% from early July through late September, but related spending rose just 19%. Tokens are the units AI uses to process data, meaning the companies surveyed paid less to process the same amount of data. Reuters Breakingviews identified declining revenue per token, driven by price competition among AI companies, as a factor behind the growth slowdown.

The company's record of profitability is also too short to conclude that it has moved beyond annual losses. Anthropic's operating loss last year exceeded USD 8 billion (approximately USD 8.16 billion), more than double the USD 3 billion (approximately USD 3.06 billion) recorded in 2024. Revenue rose from USD 400 million (approximately USD 408 million) to USD 4.6 billion (approximately USD 4.69 billion) over the same period, but annual operating expenses exceeded USD 12 billion (approximately USD 12.2 billion), widening the loss. In other words, last year it spent more than USD 2.6 (approximately USD 2.65) in operating expenses for every USD 1 (approximately USD 1.02) of revenue. It also posted an operating loss of USD 1.7 billion (approximately USD 1.73 billion) in the first quarter of this year, and its second-quarter profit offset an amount equivalent to 60% of that loss. The operating loss was about 4.3% of total first-half revenue, leaving the company in the red for the half-year.

Table 1. Anthropic's Operating Results and Profitability

Long-Term Infrastructure Commitments Set to Consume IPO Proceeds

Major expenditures are also due under computing and infrastructure contracts already signed. Commitments to use cloud computing resources from Google, Amazon and Microsoft (MS) alone total at least USD 269.4 billion (approximately USD 275 billion), according to the filing. The agreement with Google calls for payments of at least USD 114.5 billion (approximately USD 117 billion) from April of this year through July 2033. Anthropic committed to spend at least USD 110 billion (approximately USD 112 billion) with Amazon from May of this year through April 2036, and at least USD 44.9 billion (approximately USD 45.8 billion) with Microsoft from November of this year through May 2033. All three contracts require minimum spending regardless of actual usage, so the payment obligations remain even if the company cannot fully utilize the computing resources it has secured.

Beyond cloud usage fees, semiconductor and equipment leasing costs are also substantial. Lease payments under its equipment agreement with Broadcom amount to USD 161.2 billion (approximately USD 164 billion), while its agreement to use SpaceX's Nvidia graphics processing unit (GPU) computing resources is worth up to USD 120.5 billion (approximately USD 123 billion) through 2029. It has separately signed contracts to pay computing resource providers Nscale and Lambda up to USD 44.6 billion (approximately USD 45.5 billion) and USD 41.3 billion (approximately USD 42.1 billion), respectively. Payment schedules and performance conditions vary by contract, but spending needed to operate AI services—from securing semiconductors to using cloud services and leasing equipment—is committed over multiple years.

If a substantial share of the IPO proceeds goes toward payments under existing contracts, the company's capacity for additional investment after listing will shrink accordingly. As of the end of June this year, fixed and minimum payment commitments under noncancelable purchase agreements totaled USD 265.6 billion (approximately USD 271 billion), while future payments under leases that had not yet commenced reached USD 321.4 billion (approximately USD 328 billion). Some of these contracts require payments for up to 20 years regardless of whether the facilities are used, making it difficult to cut spending even if demand falls short of expectations. The Broadcom equipment lease also requires rental payments to begin before the data centers are ready, raising concerns that delays in bringing facilities online could create a cash burden. If funds are depleted fulfilling existing commitments and the facilities take longer to generate sufficient revenue, the money raised through the listing is likely to have a weaker effect on further growth and valuation gains.

Concerns Over a Post-Listing Share Price Decline

With Anthropic listing under the weight of enormous spending obligations, a high offering price would also increase the risk of a share price decline once trading begins. Art Hogan, chief market strategist at B. Riley Wealth Management, cited an excessively high offering price and a share price drop immediately after listing as his biggest concerns about Anthropic's IPO. Indeed, after this month's investor briefing, some existing and prospective investors lowered their valuation of Anthropic from USD 2 trillion to USD 1.5 trillion. That is 25% below the upper end of the expected valuation, highlighting substantial differences in investors' assessments even ahead of the listing. Even if the company secures a USD 2 trillion valuation in the offering, it remains uncertain whether buying demand in subsequent trading will support the same price.

Allowing existing shareholders to sell early in the listing is also expected to weigh on the stock. The filing outlines a plan allowing some early investors and employees to sell shares on the first trading day, followed by additional sales on the 27th and 48th trading days. Under this schedule, the supply of shares available to the market will increase as selling restrictions are lifted in stages. In particular, a surge in the share price immediately after listing would give early investors and employees a greater incentive to sell and lock in gains. Without sufficient new buying demand, profit-taking sales could accumulate and derail the rally. The prospect of shares coming onto the market could weigh on the stock even before selling restrictions expire. Investors anticipating additional supply may postpone purchases, weakening buying demand before existing shareholders begin selling in earnest.

If a sharp post-listing share price decline is accompanied by allegations of inadequate disclosure, the burden could expand to shareholder class-action lawsuits. U.S. securities law allows investors to seek damages from issuers, directors and IPO underwriters if a registration statement contains material misstatements or omissions. The central issue in such cases is whether investors were adequately informed of signs of deteriorating performance or cost burdens identified before the offering. Facebook, for example, faced a class-action lawsuit alleging that it concealed information about its growth outlook after its shares traded below the offering price for more than a year following its 2012 listing. Shareholders alleged that the company told underwriting banks to lower their earnings forecasts while withholding the information from ordinary investors. Facebook agreed to pay USD 35 million (approximately USD 35.4 million) without admitting wrongdoing, and a court approved the settlement in 2018. If Anthropic's post-listing performance falls well short of the growth outlook presented during its offering, investors could similarly question whether it adequately disclosed information needed to assess its valuation, including profitability and the scale of spending.

OpenAI Lowers Valuation Target by 11%

OpenAI's decision to postpone its listing until next year is also seen as reflecting both valuation scrutiny and fundraising pressures. OpenAI is seeking an additional USD 30 billion (approximately USD 30.6 billion) in investment in case a weaker IPO market further delays its plans. The aim is to secure funding in advance to fulfill large infrastructure contracts and sustain the company until its listing. To that end, it has discussed participation with existing shareholders, including SoftBank and Abu Dhabi investment firm MGX, and held a briefing for long-term investment funds and sovereign wealth funds in the middle of this month. OpenAI is particularly focused on securing investment commitments before Anthropic's listing. The two companies are approaching many of the same major investors, suggesting an effort to secure their money before funds are allocated to Anthropic's offering.

OpenAI has also lowered its valuation expectations to secure funding. Early this month, it proposed a valuation of about USD 1.4 trillion (approximately USD 1.43e+3 billion) to investors, but after some pushed back, it adjusted its target to USD 1.25 trillion (approximately USD 1.27e+3 billion) in recent negotiations. The figure is about 11% below the initial proposal, reflecting its acceptance of investors' demands for a lower price as it seeks additional funds ahead of a listing. It set the 30th of this month as the deadline for investors to confirm participation, while leaving open the possibility that the amount raised could fall short of its USD 30 billion target depending on demand. With the fundraising not yet complete, the final valuation and amount secured will be determined in the remaining negotiations.