We Read Anthropic's Secret Accounts!

The first real balance sheet of frontier AI

The leaked Anthropic IPO filing matters because it does something private AI markets have avoided for years: it puts numbers next to the mythology. Until now, investors could debate frontier models through demos, benchmarks, funding rounds and customer logos. The prospectus, as summarized in recent reporting and analysis, reframes the question around revenue, losses, customer concentration, infrastructure liabilities and risk factors [1].

The working headline is blunt because the document is blunt: “We Read Anthropic’s Secret Accounts!” The central story is not simply that Anthropic wants to go public. It is that one of the world’s leading model developers may become the first large pure-play frontier AI lab to ask public investors to underwrite the economics of intelligence at industrial scale [1].

The timing sharpens the stakes. Fresh market commentary says Anthropic is still being discussed as a possible Nasdaq listing before Thanksgiving, which would put the company in front of public investors before the market has had a full cycle to understand whether the AI boom is durable business formation or a capital-intensive race with uncertain margins [3].

Growth that looks historic, losses that need translation

The headline growth numbers are extraordinary. HelloBro’s briefing cites revenue of about $386 million in 2024, roughly $4.6 billion in 2025, and a much higher annualized run rate in 2026, while warning that run-rate figures are not the same as audited annual revenue [1]. TechNewsReel similarly reports that Anthropic generated approximately $4.6 billion in 2025 revenue and faced an operating loss of more than $8 billion [2].

That distinction matters because the most viral figure, a reported net loss near $42 billion, is not the same as ordinary cash burn. The leaked materials are described as including roughly $34 billion of accounting charges tied to convertible instruments, meaning the GAAP loss is scarier than the operating picture but still signals a complex capital structure built around earlier strategic backers [1].

For public investors, the question is not whether Anthropic can grow. It clearly has. The question is whether it can convert that growth into durable margins after the cost of model training, inference, safety work, customer support, sales, data-center access and hardware commitments is included. New Constructs’ analysis of the prospective offering argues that revenue growth without pricing power does not create enough value to justify the reported $2 trillion valuation target [4].

The $518 billion compute question

The most important number in the leak may not be revenue or loss. It may be the reported $518 billion of cloud, compute and infrastructure commitments over the coming years. TechNewsReel reports that about 80% of these obligations are structured as non-cancellable or “take-or-pay” commitments, with large exposure to Google, Amazon, Microsoft and Broadcom-linked equipment leases [2].

That turns Anthropic from a software story into an infrastructure story. Classic software businesses scale by selling the same code at high gross margins. Frontier AI companies scale by renting, buying or pre-committing to vast quantities of compute, then hoping customer demand, pricing and utilization arrive in the right order. If demand accelerates, those commitments look visionary. If adoption slows, prices fall, or cheaper models take share, they become a balance-sheet trap [2].

HelloBro’s own summary captures the strategic trade-off: some investors may see the commitments as manageable if revenue continues to compound, while others worry about whether enterprise customers will become more price-sensitive and route simpler tasks to cheaper models [1]. That is the core economics debate. Anthropic is betting compute scarcity remains the bottleneck. Skeptics are betting model access becomes commoditized faster than the company can build differentiated products.

Risk factors unlike a normal tech IPO

The filing is also unusual because the risk factors are not limited to familiar warnings about competition, cloud dependence or customer concentration. The leaked materials reportedly devote a large share of the document to risks from advanced AI systems, including catastrophic or existential scenarios, and warnings about models that could resist shutdown, manipulate information or behave unpredictably [1][4].

That is not just theater. Anthropic’s brand is built around safety, constitutional AI and responsible deployment. But in an IPO context, safety language becomes legally and financially charged. The company is effectively telling investors: the same technology that could justify an enormous valuation could also create harms that are difficult to predict, contain or insure against [4].

The political context is already live. Axios reported on October 5 that Senator Bernie Moreno sent a letter to CEO Dario Amodei criticizing what he called an “alarmist approach” to AI, explicitly connecting Amodei’s warnings to the expected IPO and the tension between asking Americans to invest in the company and warning that the technology could threaten them [6].

That makes Anthropic’s filing a public-policy document as much as a finance document. Public investors are not just evaluating a business. They are evaluating a company whose product may shape cybersecurity, enterprise automation, defense workflows, legal services, coding, scientific research and public anxiety about superintelligent systems.

The valuation fight has begun before the roadshow

Even before a public price range appears, the valuation argument has started. The Meridiem frames the potential Nasdaq debut as the moment when private-market AI euphoria meets public-market discipline, noting that some analysts already call the offering a test of whether frontier AI valuations can survive quarterly scrutiny [3].

New Constructs is much harsher. Its report calls Anthropic “the most ridiculous IPO of 2026,” argues that the business lacks a viable path to profit at the rumored valuation, and says the offering could function more as exit liquidity for insiders than as wealth creation for public investors [4]. That view may prove too bearish, but it highlights a real asymmetry: insiders know more about unit economics, renewal quality, compute contracts and customer concentration than future retail buyers will.

There is also a procedural caution. OrcaRouter’s October 7 explainer notes that, on the company’s own public record, Anthropic had confidentially submitted a draft S-1, but no public filing, share count, share price, listing date or final venue had been officially established by the company [5]. In other words, the leaked filing may be the best window yet into Anthropic’s business, but it is not the same thing as a final public prospectus.

Why this IPO would define more than Anthropic

If Anthropic lists successfully, it could create the first public-market benchmark for a frontier model developer. That benchmark would affect OpenAI, xAI, AI infrastructure providers, neocloud companies, semiconductor suppliers and enterprise software vendors trying to embed AI into workflows. If the deal struggles, the entire private valuation stack around AI could reprice quickly [3].

The company’s case is powerful: revenue is exploding, Claude has enterprise credibility, developers use its coding tools, and the company sits near the top of the frontier model race [1]. But the countercase is equally clear: massive compute obligations, falling model prices, growing open-source competition, customer concentration and legal exposure create a business model that may not resemble the high-margin software platforms investors know [2][4].

The leaked accounts therefore read like Skynet’s fiscal version finally arriving with an audit trail. They show an AI lab that is both stronger and riskier than the hype suggests. Stronger, because the revenue curve is real enough to force Wall Street to pay attention. Riskier, because the cost of staying at the frontier may be so high that growth alone is not proof of value.

The public markets are about to ask a colder question than the venture market did: not “Can Anthropic build astonishing models?” but “Can astonishing models become a profitable public company?” That is the secret account that matters most.

Sources from the last 72 hours

1. [1]We Read Anthropic's Secret Accounts!Oct 7, 2026, 6:00 PM
2. [2]Anthropic commits $518 billion to AI infrastructure in massive compute betOct 7, 2026, 2:00 AM
3. [3]Anthropic IPO Tests AI Valuations as Public Markets Brace for RealityOct 8, 2026, 2:11 AM
4. [4]Anthropic Is The Most Ridiculous IPO of 2026Oct 6, 2026, 2:00 PM
5. [5]An Anthropic IPO: What Is Established, What Is Speculation, and What Would Have To Happen FirstOct 7, 2026, 2:00 AM
6. [6]GOP senator warns Anthropic of "alarmist" approach to AIOct 5, 2026, 8:00 PM

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.