Anthropic faces immense competitive pressure and substantial fixed costs. Can Claude accelerate its development fast enough to ensure success?
The Anthropic S-1 filing appears to provide the financial justification for the company to invest heavily in compute. For those wondering if Anthropic is attempting to construct a "machine god" while convincing Wall Street it sells IT contracts, the filing offers a solution: actions that resemble building a "machine god" are the only way to avoid bankruptcy.
Set aside the "machine god" concept for a moment. According to my reading of the S-1, Anthropic pins its future on dramatic growth; it has no choice given its fixed costs. While one does not need to be an AI skeptic to question these figures, a skepticism regarding dramatic recursive self-improvement—essential to the business plan—suggests this may not be a prudent investment.
Good news for Anthropic supporters includes revenue growth fourteenfold to $16.3 billion between the first half of 2026 and the first half of 2025. Despite an $80 billion loss, largely an accounting figure, adjusted income reached $1.62 billion when excluding certain charges, including employee compensation. The S-1 suggests that, at least for this company right now, AI can generate enough revenue to pay for itself.
Anthropic names only OpenAI and Google as rivals. The company was more disciplined than OpenAI or Elon Musk, a strategy that seems to be paying off. Anthropic refrained from releasing AI image generators, TikTok clones, or waifu-focused tools. Instead, it specialized in tasks like coding and cybersecurity, focusing on computer-centric tasks. It targeted the enterprise market, which is less fickle than consumers, resulting in a net revenue retention of 900 percent. Over three-quarters of customers spend at least $100,000 annually.
However, Anthropic faces intense pressure from both frontier labs and open-source models. The other labs are racing for the cutting edge; Claude Opus 5.5 currently sits atop the Epoch Capabilities Index, though rankings fluctuate. Meanwhile, open-source models offer capabilities four months behind the latest releases.
Furthermore, Anthropic has a disadvantage against a rival like Google, which has multiple revenue paths. Google can integrate AI into existing business products. Microsoft Teams' success against Slack illustrates that distribution matters. While the best model is critical for cybersecurity—Anthropic's growth target—competing on distribution is difficult.
Long-term commitments highlight the immense expansion pressure on Anthropic. While specific figures on future spend are partially redacted, known commitments involve a fixed minimum of $265.6 billion in purchase commitments and $321.4 billion in unused lease agreements for compute infrastructure. New agreements add another $245 billion, with other pending transactions estimated between $125 billion and $150 billion. Total minimum commitments over 20 years approach $850 billion, potentially rising to $1 trillion.
To meet these obligations, Anthropic's growth must continue at breakneck speed. Analyst Harrison Rolfes of PitchBook models annual payments at $90 billion from 2027 to 2030, requiring $188 billion in revenue to maintain the current cost-to-compute ratio. "Growth is not optional for this company," Rolfes writes. "The contracts assume it."
Compute and infrastructure costs for customers were $2.3 billion in the first half of the year, while model development cost $9.7 billion. However, Anthropic has become more efficient, spending less to generate each dollar of revenue. Analyst Brendan Burke notes the S-1 freezes in a period of high usage and scarcity, justifying higher costs.
As enterprise software, Anthropic targets IT budgets. Cybersecurity requires constant updates to stay secure, similar to Anthropic's need to constantly train. AI introduces new threats, and product headaches arise from security protocol failures. Frontier labs need high-margin products to avoid token price wars. Opus 5.5 reduced customer costs by 40%, allowing Anthropic to preserve margins through increased usage.
Gil Luria, an analyst at DA Davidson, suggests increased efficiency doesn't always harm margins, citing AWS's model. Yet, the pressure for growth is undeniable. This leads to "recursive self-improvement." Anthropic warns that it is locked into this approach, which carries a trade-off: using AI to build more AI may cause capabilities to advance faster than safety and alignment measures can be developed. Essentially, for business reasons, Anthropic may be forced to take risks.
The risk factors mention existential hazards and "catastrophic harm." Whether recursive self-improvement without human supervision is possible is unknown. If not, meeting commitments and turning a profit will be much harder.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.