Anthropic S-1: of its $81B net loss in H1 2026, $72B+ was related to rising convertible note value as its valuation increased and nearly $9B to compute

The cost of accelerating AI development through self-improvement now appears on company balance sheets.

Anthropic's S-1 filing reveals that over $81 billion of its net loss in the first half of 2026 is attributed to the rising value of its convertible notes, with more than $72 billion of that increase stemming from its escalating valuation. An additional nearly $9 billion of the loss is tied to computational costs. The filing extensively details the "existential risks" associated with advanced AI, particularly through recursive self-improvement (RSI). Anthropic acknowledges that AI systems automating research and development could accelerate progress beyond human ability to safeguard models. It expresses concerns that highly autonomous AI could take harmful actions at scale without human intervention, and when AI directs physical equipment, errors or misuse could lead to immediate and irreversible damage. Despite these risks, Anthropic views "closing the loop" through RSI as a critical "structural advantage" in frontier AI, fearing falling behind competitors if it does not pursue it.

The significant financial reporting in Anthropic's S-1, detailing losses tied to convertible note valuation and compute, underscores the immense capital investment and financial maneuvering in the frontier AI race. This highlights how current AI development, even with its theoretical risks, is deeply intertwined with financial market valuations and investment strategies, suggesting a landscape where theoretical dangers are being actively pursued for competitive and financial gain.

Anthropic's dual framing of recursive self-improvement as both an existential risk and a core competitive advantage reveals a fundamental tension within the AI industry. This dichotomy suggests that despite industry-wide discussions on AI safety, the drive for advancement and market position may inherently push companies toward the very capabilities they deem potentially catastrophic, creating a precarious balancing act for all involved.