Anthropic filed its IPO registration statement last week, targeting a November debut with topline numbers that look impressive for a startup — fast revenue growth, $96 billion in cash reserves. The surface financials tell one story. The footnotes tell another. Buried in the amended registration statement: at least $582 billion in non-cancelable spending commitments to five companies — Amazon, AMD, Broadcom, Google, and Microsoft — for the compute infrastructure required to train and run its models. Total contractual commitments reach $895 billion, most stretching into the 2030s, and the company is still negotiating more. These are legally binding future payments that do not appear on the balance sheet. Anthropic could cross $1 trillion in commitments before year-end. The revenue engine needed to service those obligations is sputtering. Third-quarter 2026 results suggest growth slowed dramatically over summer. The company's 2026 financials have not been formally audited. Its much-hyped Fable 5 frontier model received a tepid commercial reception. A study found the share of Americans using AI at work dropped from 46% (summer 2025) to 33% (summer 2026). The demand story is not cooperating. Four structural forces are squeezing margins simultaneously. Demand uncertainty: businesses aren't demonstrating willingness to pay for ever-larger models. Commoditization: Google, Meta, Microsoft, OpenAI, SpaceX, and Chinese open-weights competitors offer largely interchangeable products, triggering a price war between Anthropic and OpenAI this summer. Cost pressure: a single 1 GW data center costs an estimated $31 billion — Anthropic has committed to 15 GW and estimates needing 20-25 GW. External drag: community opposition has scuttled data center projects, the Iran War is disrupting supply chains, and construction delays limit the capacity Anthropic is counting on for revenue. The systemic risk argument is straightforward. Public markets are already historically overvalued and concentrated in a handful of interconnected tech mega-caps. Adding another massive, unprofitable tech company increases contagion risk. As a private company, Anthropic's losses fall on corporate insiders, tech giants, and VCs. Once shares enter 401(k)s, index funds, and retirement accounts, ordinary workers and retirees absorb the downside when the margin squeeze bites. The SEC has statutory authority to delay or block an IPO if the registration statement is "on its face incomplete or inaccurate in any material respect." The author argues that Anthropic's unaudited financials, off-balance-sheet obligations dwarfing its cash reserves by nearly 10:1, and slowing growth amid a structurally hostile competitive environment meet that threshold. The question is whether the agency will use that authority or wave through a filing that transfers catastrophic risk from Silicon Valley insiders to the public. This is not an abstract debate about AI's future. It is a concrete question about who pays when $895 billion in commitments collide with a business model that cannot yet generate the revenue to cover them. The answer, absent intervention, is every American with a retirement account.
