Rybo

Spoiler alert! I read the Anthropic S-1, and I'd recommend everybody do their own reading! The TLDR upfront: No surprise. This is not a company that is run for the benefit of shareholders. It reads to me that shareholders of Anthropic are almost like owners of Bitcoin in a way. It's a science experiment that goes on and will have a price attached to it, not exactly a claim on cash flows run by stewards interested in maximising those cash flows for the benefit of owners. The income statement is fantastic and wildly misleading. The TL is my notes + questions from the S-1 based on my initial 2hr long read: 1. The letter/pitch at the start of the S-1 is very hippie-esque. Not "Elevating The World's Consciousness" level like Adam Neumann's WeWork, but it's up there. Equity match for philanthropic contributions, equity kept aside for the public good, and more to be done in lieu of executive compensation, so expect dilution in line with other public companies. I'm like, "these guys don't believe in carrots and sticks", at least on an individual level. This is effective altruism as an S-1. 2. Dual-class shares in the capital structure, with the founders controlling >50.1% of the votes. Giving that much power to people that don't sound like they care about market capitalisation or equity value per share feels dangerous. The LTBT elects a board majority through financially disinterested trustees. 3. Anthropic is a brilliant company from a product perspective. Put Frank Slootman in charge of this thing, and it'll crush everything in sight while making money for investors. 4. The pro-forma metrics are wild! Not Community-Adjusted EBITDA, but wild. How are these not operating expenses or COGS? Vendor costs related to data quality and performance evaluation services, information technology and cloud software costs, payment processing and platform fees, facility and occupancy costs, non-income taxes, professional services costs, bad debt expense, and depreciation and amortization related to non-data centre assets. We should call Anthropic's Adjusted Operating Income as Constitution-Adjusted EBITDA. 5. Compute and Infrastructure expense as a percentage of revenue dropped, but how much did the compute the company is paying for, in megawatts, expand? If Anthropic grew its revenue 13x while compute it was paying for grew 3x, that is impressive, but it is impossible to tell if the forward-looking compute arrangements it is paying for will translate into incremental revenue at the same pace. 6. What are these "Change in fair value of certain mark-to-market financial instruments" that are somehow making this company "adjusted net income" profitable? 7. So they've got note purchase agreements and similar arrangements with Amazon, Google, Nvidia, AMD and Broadcom to finance their cloud/GPU bills with Anthropic stock. I've never seen this before. 8. $390B in combined committed infrastructure spending to Microsoft, Google, Amazon, and SpaceX between 2026 and 2033-2036, depending on the contract. 9. There's an expectation to spend $20B on AMD racks over an undefined period of time. 10. There's $45B on Nscale for Nvidia GPUs and $35B on Lambda for the same by 2033-2034. 11. $161B in Broadcom/Google TPU lease rent payments over an unspecified length of time. 12. What is the revenue growth required to be able to service those bills? 13. I'm not sure there is enough equity and debt capacity in the biggest capital market in the world for this company. Elon will need to get us to Mars and help Dario sell Anthropic securities to Martians so the company can pay its bills. Conclusion: Either I'm going to be wildly wrong, this company continues to grow 1300% YoY for the next 5-6yrs and everything is going to exceed my expectations, or, when either Anthropic or OpenAI run into trouble, we are all in for a world of pain from the largest restructuring(s) in history.