• Anthropic warns in its confidential IPO filing that government attitudes toward the company could materially harm its business, finances, and market position.
  • The Pentagon’s supply-chain risk designation, upheld by a divided appeals court, has already cost Anthropic military contracts and billions in potential revenue.
  • The company’s planned public debut may slip past the November midterms as it navigates litigation, safety restrictions, and massive infrastructure commitments.

A Safety-First Stance Meets National-Security Realities

Anthropic’s pursuit of a record-scale IPO is colliding with an increasingly hostile government posture. According to a Reuters review of the company’s confidential IPO materials, the AI developer explicitly cautions that “government attitudes toward Anthropic and its technology—including policy, procurement, and regulatory decisions—could adversely affect operations, finances, and market position.” The warning is unusual in its specificity, and it underscores a central tension: a company built on safety-first principles is now fighting for access to the world’s largest customer.

The most immediate flashpoint is the Pentagon dispute. In March, the Defense Department designated Anthropic a national-security “supply-chain risk” after the company refused to permit uses involving autonomous weapons and mass surveillance. On September 25, a divided U.S. appeals-court panel upheld that designation, though a separate California federal-court ruling had blocked aspects of broader administration-wide restrictions. Anthropic says it disagrees with the ruling and is considering further appellate review.

The government conflict is particularly consequential ahead of a public listing, as investors must weigh not only growth prospects but also exposure to federal procurement, policy shifts, litigation, and reputational fallout. Reuters reports that the planned IPO may be delayed until after the November U.S. midterm elections.

Billions at Stake as Military Contracts Slip Away

The financial impact is already being felt. Anthropic says the designation has restricted military-contract access and cost it billions in business and reputational damage. That is a significant admission for a company with a reported annualized revenue run rate exceeding $65 billion by the end of July 2026, following a May 2026 post-money valuation of $965 billion.

The company’s 2025 performance illustrates both its explosive growth and its heavy losses: revenue rose roughly 12-fold to nearly $4.6 billion, but operating losses exceeded $8 billion. Reuters also reported a $42 billion net loss for 2025, reflecting the enormous cost of frontier-model development and infrastructure. To sustain that growth, Anthropic has disclosed at least $518 billion of planned AI-infrastructure commitments over a decade, about 80% of which are non-cancelable or payable whether capacity is used or not. Minimum commitments include $111.1 billion to Google (GOOG), $110 billion to Amazon (AMZN), and $31.4 billion to Microsoft (MSFT), alongside large Broadcom (AVGO)-related lease obligations.

Those obligations loom over the IPO narrative. “What institutional investors like us are really focused on is regulatory stability,” said one person familiar with the matter, echoing concerns that policy risk could undermine Anthropic’s ability to service its fixed costs.

A Governance Structure That Limits Shareholder Power

Beyond government relations, Anthropic’s governance plan is raising eyebrows. The company intends to establish a “Founder LLC,” controlled by its seven co-founders, holding Class F voting power equal to 50.1% over key matters. It will also remain a Delaware Public Benefit Corporation. The prospectus acknowledges this could lead to decisions that conflict with shareholders’ short-, medium-, or long-term financial interests—a notable caveat for public investors.

There has been no Reuters-reported change of CEO or top leadership. Co-founder Dario Amodei remains CEO, with Daniela Amodei also among the founders. The company’s Claude family of generative-AI models competes with systems from OpenAI, Google, xAI, and others, and Anthropic earns revenue largely through metered usage and subscriptions, often sold through or alongside major technology partners.

Partner Concentration and Competitive Pressure

Big Tech partners account for 47% of 2025 revenue and are simultaneously investors, customers, distributors, compute suppliers, and AI competitors. That concentration creates meaningful conflict-of-interest risk, and it underscores the sector-wide shift toward capital intensity. Anthropic is moving from a primarily cloud-based approach toward dedicated infrastructure and directly leased chips, which improves control over scarce capacity but transfers enormous fixed-cost and utilization risk onto the company.

The broader AI investment cycle is increasingly defined by financing capacity and access to specialized hardware. Anthropic’s $518 billion buildout is comparable in headline scale to OpenAI’s reported $500 billion Stargate plan. Compute availability, not customer demand, is likely to be the primary constraint on future advanced-AI supply, according to the company.

Political and International Fallout

The central political issue is a clash between AI safety guardrails and the government’s desire to use advanced AI for broad national-security purposes. Anthropic sought to prohibit use of Claude for fully autonomous lethal weapons and mass surveillance of Americans; the Pentagon sought broader permission for lawful military uses, leading to the supply-chain-risk designation after negotiations broke down. The appellate court found the designation reasonable and rejected Anthropic’s claim that it was retaliation for its AI-safety views.

Internationally, Anthropic warns that large-scale AI could concentrate wealth and power and destabilize the geopolitical order. CEO Dario Amodei has urged international cooperation and a slower, safety-conscious pace of frontier-model deployment. Reuters reports that he and leaders from Google DeepMind, OpenAI, and xAI recently called for a slowdown to focus on safety. That stance, however, is in tension with competitive pressure: Anthropic has considered releasing a model to counter momentum from OpenAI’s GPT-6 Astra.

Litigation and Legal Risks Mount

Copyright litigation adds another layer of risk. Anthropic obtained approval for a $1.5 billion settlement with authors in July, but subsequently faced an August lawsuit from Round Hill Music alleging misuse of songs to train AI systems. These disputes add legal and reputational exposure around training data, even as the company devotes roughly 80 pages of its 261-page main prospectus body to risk factors—far more than the 48 pages focused on the business itself.

The company warns that models could show self-preserving behavior, resist shutdown, conceal or manipulate information, or generate behavior resembling blackmail. Such disclosures underscore the unique societal and regulatory scrutiny facing frontier AI developers.

Outlook: A High-Stakes Balancing Act

In the short term, the Pentagon designation remains an overhang unless Anthropic wins rehearing or a later reversal. The company’s ability to sell directly or indirectly into defense-related ecosystems, and the signaling effect for other federal customers, will be important for its IPO narrative. Investors will also scrutinize whether the reported $65 billion annualized revenue run rate can sustain utilization of its immense infrastructure commitments.

Longer term, if Anthropic maintains technological leadership and enterprise demand keeps growing, its infrastructure investments could secure scarce compute and reinforce its competitive position. The company projects roughly $190 billion–$200 billion in 2028 revenue, though that is an internal forecast reported by Reuters rather than an externally verified outcome. Conversely, political conflict could limit government sales, make commercial partners cautious, invite tougher rules, or alter access to critical infrastructure and chips.

The high-control founder structure may protect Anthropic’s safety mission during commercial pressure, but it also creates a durable governance tradeoff: public shareholders may have limited ability to change strategy even if costs, policy friction, or safety controversies intensify. Analysts see a valuation challenge. Reuters Breakingviews characterized a proposed $2 trillion valuation as highly ambitious given large losses and capital obligations, even amid unusually fast revenue growth and a reported 40% gross margin in 2025.

Anthropic declined to comment beyond its public filings. A Pentagon spokesperson did not respond to a request for comment.

Correction: An earlier version of this article misstated the date of the appeals-court ruling. It was September 25, not September 24.