NEW YORK — Anthropic, the developer of the Claude artificial intelligence model, has begun work in earnest on a Nasdaq listing. With AI-related stocks caught up in a debate over excessive investment following the credit downgrade of CoreWeave, the world's largest provider of graphics processing unit infrastructure, Anthropic's initial public offering is expected to serve as a weather vane for the stock market heading into year-end. Weakening investor sentiment toward AI infrastructure has trimmed Anthropic's estimated valuation from the $2 trillion (about 2,700 trillion won) figure once discussed. The company drew attention by filling its IPO application with the warnings about human extinction that Chief Executive Dario Amodei has recently voiced. Still, if Anthropic succeeds in raising a large sum on the strength of its blueprint for the AI era, that could offer hope for the IPOs of other privately held AI companies, including ChatGPT developer OpenAI. Depending on the valuation Anthropic commands, the market capitalization rankings among technology stocks could be shaken up, as they were when SpaceX went public on June 12 of this year.
Anthropic Files Registration Statement, Targeting Valuation of Up to $2 Trillion and $100 Billion Raise
The Dow Jones Industrial Average (0.28%), the Standard & Poor's 500 (0.70%) and the Nasdaq Composite (0.83%) all closed higher in New York on the 26th. The Philadelphia Semiconductor Index also jumped 2.82%, its first strong gain in some time.
The rise in New York stocks, led by technology shares, came on news that Anthropic had filed its Nasdaq IPO application with the U.S. Securities and Exchange Commission that day. Anthropic had earlier submitted a draft registration statement to the SEC on June 1. At that time the company had expected to list this month, but it pushed the timing back to after the Nov. 3 midterm elections as investment conditions deteriorated in the second half, with U.S. Treasury yields surging and doubts spreading about AI profitability.
Anthropic's debut on the U.S. market would be the first by an AI model company. Goldman Sachs and Morgan Stanley are serving as joint lead underwriters. Anthropic will list Class A common shares on the Nasdaq next month under the ticker symbol ANT. The registration statement did not disclose the number of shares to be offered or a proposed price range. The Wall Street Journal reported that Anthropic is aiming to be valued at as much as $2 trillion in the IPO and to raise $100 billion (about 135 trillion won). That would be the largest on record, exceeding both the $1.77 trillion valuation and the $75 billion raise that SpaceX achieved when it listed on the Nasdaq on June 12.
Anthropic posted revenue of $16.286 billion (about 21.6 trillion won) in the first half of this year, more than a 14-fold increase from $1.16 billion in the same period last year. Annual revenue last year was $4.59 billion, up 1,089% from 2024. Its net loss came to $80.748 billion (about 107 trillion won). That figure reflects about $72.4 billion in non-cash accounting charges tied to changes in the value of convertible bonds and derivatives issued during earlier fundraising. Excluding the one-time accounting valuation losses, adjusted operating profit was $1.618 billion (about 2.1 trillion won), the first half-year operating profit in the company's history. As of June 30, cash and cash equivalents stood at about $112.4 billion and short-term investments at about $19.3 billion.
Corporate customers generating annual revenue of $100,000 or more surged to 11,900 in the first half of this year from 950 in the first half of last year. As of Sept. 30, customers generating annual revenue of $1 million or more also exceeded 1,900. About 250 customers generated annual revenue of $10 million or more over the past 12 months.
The filing also disclosed Anthropic's unusual governance structure. The company granted a single Class F share to Founder LLC, a separate limited liability company set up by its co-founders. The share is designed to carry 50.1% of the voting power on general shareholder-meeting items. That allows the founders to control most agenda items regardless of their actual stakes. At the same time, a Long-Term Benefit Trust made up of five independent trustees holds the right to elect and remove a majority of directors through Class T shares, preventing the co-founders from dictating the composition of the board.
Strategic partners that previously invested in Anthropic, including Amazon and Google, were given only economic rights such as dividends. They received only non-voting Class N shares, leaving them unable to influence control of the company. Class N shares convert into Class A shares only when sold to a third party.
The co-founders, who each hold between 230 million and 250 million shares, have pledged to donate 80% of their personal stakes to philanthropy. At the corporate level, Anthropic has also contributed 5% of its total shares to a public-benefit fund.
A Quarter of the Filing Is Devoted to AI Risks, Along With the Memory Supply Chain, Friction With the Trump Administration and Climate Change
Anthropic took the unusual step of describing the dangers of AI in detail in its registration statement. Risk factors ran to 94 pages of the roughly 400-page document, about a quarter of the total. "If our systems are misused, replicated, weaponized or operated with a high degree of autonomy, they could pose catastrophic risks that threaten human existence," the company said. It also flagged as a major risk factor its current legal dispute with the U.S. Department of Defense, also known as the Department of War. The reason cited was the possibility that companies selling products and services to the administration of President Donald Trump would stop using Anthropic's AI.
The filing warned in particular about so-called alignment failures, in which AI behaves in ways inconsistent with the purposes, values and ethical standards intended by humans and causes unexpected harm. Anthropic reported observing cases in which AI models resisted shutdown in testing and demonstration environments, or concealed and manipulated information and resorted to coercion and deception. It also included concerns that malicious human actors could abuse AI models to develop chemical, biological, radiological and nuclear weapons. As a prominent example, it cited an incident in September of last year in which a hacking group believed to be backed by the Chinese government manipulated Anthropic's Claude Code in an attempt to infiltrate more than 30 targets, including large corporations and government agencies worldwide.
Anthropic stressed that while it is deploying technical safeguards to detect and block such misuse, there is no guarantee they are complete. It also stated that the liability and legal exposure it would face in the event of an AI misuse incident would be enormous, and that insurance to cover it may be difficult to obtain.
Anthropic described supply chain uncertainty, including for memory chips, in detail as a risk factor. Above all, it said the concentration of high-bandwidth memory supply — a core component of AI accelerators — among three companies, Samsung Electronics (005930.KS), SK hynix (000660.KS) and Micron, could hurt its ability to secure advanced chips. It wrote that most of their production facilities are outside the United States and could be affected by trade controls, sanctions and tariffs. It added that manufacturing of AI accelerators is overwhelmingly concentrated at Taiwan's TSMC, meaning that friction across the Taiwan Strait, a military clash or a natural disaster could paralyze global semiconductor supply for years. Anthropic's minimum payment obligations to secure computing capacity totaled $245 billion.
The company also raised water depletion driven by the environment and climate change, and difficulty securing cooling water, as problems. Its list of concerns included local opposition across the United States to data center construction. Bloomberg noted that SpaceX included none of the environmental and climate concerns raised by Anthropic in its own registration statement.
Anthropic's filing did contain more positive assertions about AI than negative ones. The company projected that AI, if used appropriately under human control, would deliver enormous economic value to humanity. It forecast that once advanced AI reaches the stage of "powerful AI" that surpasses the capabilities of human experts, scientific progress that would take decades could be compressed into a few years. It also cited projections from research firm IDC that AI will sharply raise the productivity of knowledge workers, generating a cumulative economic impact of $22.5 trillion (about 30,000 trillion won) by 2031. "AI could play a decisive role in potentially curing diseases that have afflicted humanity for thousands of years, including cancer, Alzheimer's, heart disease, rare genetic disorders and major infectious diseases," Anthropic said.
Early Investors Stand to Profit, but Souring Sentiment Could Pull the Valuation Down to $1.5 Trillion
Bloomberg assessed that Anthropic's listing will deliver enormous gains even to later-stage investors. Anthropic's seven co-founders are Amodei; his sister Daniela Amodei, the president; Tom Brown, chief compute officer; Jared Kaplan, chief science officer; Sam McCandlish, chief architect; Chris Olah, who leads interpretability research; and Jack Clark, who heads policy research. Bloomberg named Iconiq, which first invested more than $6 billion in Anthropic in September of last year, among the later investors set to make large sums. At that time Anthropic's valuation was just $183 billion, far below OpenAI's. If Anthropic is valued at $2 trillion, the value of Iconiq's stake would rise to $34 billion.
Spark Capital, the only venture capital firm to secure a seat on Anthropic's board, invested $75 million in a 2023 round that valued the company at $5 billion. That stake is now worth $7 billion. Sound Ventures, the venture capital firm founded by prominent manager Guy Oseary and actor Ashton Kutcher, also scored a major success, starting with $35 million in the same round and investing $400 million in total.
Sequoia Capital, the well-known Silicon Valley venture firm, made its first investment of $3 billion in Anthropic earlier this year, securing a stake of about 1%. Singapore's sovereign wealth fund GIC has also invested more than $4.5 billion in total since its first investment last year. Google and Amazon, both major customers, and partner Nvidia have also invested billions of dollars in Anthropic. Former Google CEO Eric Schmidt and FTX co-founder Sam Bankman-Fried took part as early investors.
Views on Anthropic's IPO are not uniformly positive. Its valuation initially surged to $965 billion in May of this year from $183 billion in September of last year. That surpassed the $852 billion valuation that rival OpenAI recorded in March of this year.
The picture then reversed as interest rates surged in the second half and market concerns about AI-related stocks grew. According to Reuters, some investors at a pre-IPO briefing held at Anthropic's San Francisco headquarters on the 14th of this month were disappointed by a slowdown in the company's growth and discussed a valuation of around $1.5 trillion (about 2,000 trillion won), far below the $2 trillion originally expected.
OpenAI has pushed its listing back to next year or later. The Financial Times reported on the 22nd that OpenAI had sought a final round of funding targeting $1.4 trillion but lowered that to $1.25 trillion after opposition from some investors. OpenAI is reported to be seeking to raise $30 billion in the round ahead of an IPO next year. It has asked investors to decide whether to participate by the 30th.
U.S. technology outlet The Information also reported that SB Energy, an OpenAI partner and data center developer under Japan's SoftBank Group, is considering delaying its U.S. listing to December or next year. SB Energy is pursuing a data center project in Ohio for OpenAI's use and had originally planned to list in September. It filed an IPO prospectus on the 2nd of last month and an amended prospectus on the 21st. SB Energy had intended to raise $5 billion to $7 billion (about 6.6 trillion to 9.3 trillion won) at a valuation of about $60 billion (about 79.3 trillion won). The complication arose from the fact that not one of its data centers has begun operating. Contracts tied to the $430 billion Ohio data center project that SB Energy has secured account for most of the company's expected revenue. According to sources, SB Energy executives therefore plan to decide whether to pursue a December listing based on how Anthropic's IPO goes.
Intensifying Industry Competition and Banks Blocking AI Agents Add to the Problems; Listing Seen as a Gauge for a Year-End Rally
Souring investment sentiment in financial markets is not Anthropic's only obstacle. Various concerns the company itself raised in its registration statement are already materializing in several places. According to The Information on the 24th, OpenAI has recently been winning away Anthropic's customers with steep discounts. At human resources software company Rippling, engineers used Anthropic models for 57% of their coding work in May but have now cut that to 18%. They raised their use of OpenAI models to 52% from 16%. The reason was cost savings.
Coding app company Replit has also switched to OpenAI products in recent months, because OpenAI cut its usage fees by more than 30% on contracts worth more than $100 million a year. AI-based service management company Atomicwork, which previously used only Anthropic models, has recently begun using OpenAI models alongside them. On the 22nd, Anthropic unveiled an AI agent called Claude Duet, following Meta's Muse and OpenAI's Dots. Claude Duet is based on Claude Opus 5.5, Anthropic's second-tier AI model. The agent monitors connected documents and apps on its own and identifies and carries out necessary tasks without being instructed step by step. Intensifying competition in the AI model market adds to uncertainty around Anthropic's growth rate, which has already clearly slowed.
Political friction is another worry surrounding Anthropic. The same day, U.S. online outlet Axios cited federal lawmakers and White House aides in concluding that current AI company chiefs are carrying a heavy political burden. "Even administration officials and lawmakers friendly to AI want the founders to keep quiet and make their products safe, like the heads of other established industries," Axios reported. The point was that unease is widespread in Washington political circles over Amodei and others continuing to raise the prospect of AI destroying humanity. Amodei caused a stir inside and outside the industry when he argued on his blog on the 12th of last month that the pace of improvement in AI model capabilities should be slowed. Unusually, Amodei's post drew agreement from rivals including OpenAI CEO Sam Altman, xAI founder and SpaceX CEO Elon Musk, and Demis Hassabis, chief scientist at Google parent Alphabet. "The AI companies' efforts have so far failed to win public opinion, and recent polling showed a majority of Americans consider AI harmful to society," Axios wrote. "The clear lesson for AI company CEOs is to talk less."
Axios also reported on the 23rd that major U.S. banks including JPMorgan have detected cases of AI agents accessing and operating in customer accounts and are tightening security measures such as blocking access. The assessment was that access bypassing bank security verification is increasing as customers hand account login credentials to AI agents for convenience. Axios added that there were also signs that some agents sought to act independently of their users' instructions. JPMorgan CEO Jamie Dimon warned in a speech at an Economic Club of Chicago event on the 14th that people assume AI will act appropriately on their behalf, but it may not.
However Anthropic's listing turns out, it is expected to be a major inflection point for global financial markets. It is no exaggeration to say that whether the AI-driven market that has run since last year extends through the end of this year and into next, or instead pauses for a considerable stretch, depends on how the market responds to Anthropic. Before that, earnings from major technology companies — Alphabet and Microsoft on the 28th, Amazon and SanDisk on the 29th — could move AI-related share prices.
Trump Stocker is a column delivering on-the-ground reporting and analysis of U.S. markets, companies, policy, politics and diplomacy that may help investors during the era of President Donald Trump. Subscribe to receive useful news from the United States.