Anthropic's business is insane
Anthropic’s growth and heavy compute commitments make its outlook depend on continued model leadership, not revenue growth alone.
- Track whether Anthropic’s revenue growth can support its take-or-pay compute contracts; the commitments remain payable even if capacity goes unused.
- Watch for cheaper competitors or a plateau in model capability: either could push customers toward lower-priced alternatives.
- Treat the S-1’s governance, loss disclosures, and insider-selling rules as risks to examine—not proof of future performance.
The video is mainly commentary and an interview; its risks and arguments are covered in the notes. · Best for: Entrepreneurs and operators assessing AI-market competition and infrastructure risk.
The creator discusses Anthropic’s S-1, focusing on slowing revenue growth, long-term compute obligations, model competition, and governance. A stock-market commentator then offers a more skeptical view of the company’s financial arrangements and potential IPO risks. The speakers’ claims and opinions are not independently verified in the transcript.
Revenue growth and compute commitments1:58
- The creator says Anthropic’s reported revenue growth slowed from Q2 to Q3, while noting it remained positive.
- The updated S-1 reportedly lists $900 billion in committed spending over about 19 years.
- The creator highlights take-or-pay terms: Anthropic may owe payment even if it does not use the contracted compute.
- The creator attributes some slower growth to customers reducing token-heavy usage, but argues Anthropic still grew after that behavior eased.
Competition, pricing, and the company’s risks5:23
- The creator sees the main risk as Anthropic losing its model-quality lead; customers may switch when another model is as capable and cheaper.
- A second risk is a plateau in the level of intelligence customers need, which could make price more important than further capability gains.
- The video raises cheaper Chinese and open-source models as potential competitive pressure.
- The creator criticizes a reported $1.5 billion settlement over books used to train models and questions the company’s governance structure.
Stock commentator: revenue, accounting, and contracts12:51
- The guest initially speculates that missing Q3 figures could reflect a desire to avoid showing slower growth; the creator notes that an updated S-1 later supplied results.
- The guest argues that investments, cloud contracts, and vendor relationships can create circular flows of money among AI companies and infrastructure providers.
- The guest cautions that revenue and accounting figures do not by themselves establish sustainable profitability, and points to the S-1’s warning that losses may continue.
- The guest sees government contracts as valuable but says multiple AI and infrastructure companies may not all be able to repay their obligations.
Governance and voting control30:15
- The guest describes the public-benefit structure and long-term benefit trust as unusual-looking but says founder voting control is not unprecedented.
- The guest’s interpretation is that layered boards and entities can obscure who makes particular decisions and spread accountability.
- The guest compares this structure with founder control and share-class arrangements at other companies.
IPO mechanics and broader market concerns37:16
- The guest flags the S-1’s disclosure of past net losses and the possibility that profitability may not be achieved or sustained.
- The guest raises concern about insider selling at IPO and rule changes that, in the guest’s view, could expose index-linked investors to a new listing.
- The discussion broadens into the guest’s views on market concentration, borrowing, and margin debt; these are not presented as a detailed Anthropic analysis.
Tools & resources mentioned
- Anthropic S-1 — filing discussed for revenue, spending commitments, losses, governance, and IPO terms.
- SEC.gov — cited by the guest as the source for the S-1.
Notes are AI-generated from the video's transcript and may miss nuance.