Anthropic diligence pack

What they do

• AI safety lab building the Claude models; runs on TPUs, Trainium, and GPUs.
• Primary segment: AI & Data Center / AI Labs / Model Builders.
• Key products in the dataset: Claude models (multi-chip trained: Trainium/TPU/GPU), Project Rainier (AWS Trainium2 compute cluster).
• Market position in that segment: major.

Role in the chain

• Anthropic is a market major in AI & Data Center (AI Labs / Model Builders), headquartered in United States. AI safety lab building the Claude models; runs on TPUs, Trainium, and GPUs.
• Ask: On the step you actually buy, is this company the process owner, a second source, or a distributor? The segment label is not a substitute for that answer.

Counterparties on the graph

• Tracked suppliers: 5. Tracked customers: 2.
• Named suppliers (first 5): CoreWeave, NVIDIA, Alphabet (Google), Amazon (Annapurna Labs), Microsoft.
• Named customers (first 2): Microsoft, Perplexity AI.
• Ask: Which of those edges are in a filing or a primary-source quote, and which are segment-inferred? Do not treat the two as the same fact.

Concentration and replaceability

Operator section

• 2 of the tracked suppliers are a monopoly or leader: CoreWeave, NVIDIA.
• Ask: If the named monopoly or leader supplier stopped shipping, is there a qualified second source on that process step? If the dataset does not name one, do not assume one exists.

Geography and jurisdiction

• Headquarters country in the dataset: United States.
• Ask: Where do the fabs, plants or logistics nodes actually sit, versus the HQ country on this record? This dataset records headquarters, not every site.

Public-markets facts (only what is on the record)

Investor section

• Private company in this dataset: no ticker, no market cap. Do not invent a valuation.
• Ask: What does the latest reported period (if any) say about capacity, backlog and customer concentration? If the company page has no revenue figure, that number is not in this dataset.

Operator checklist

Operator section

• Ask: The dataset records 5 tracked suppliers for Anthropic. Which of those edges are researched (named in a filing or primary source) versus segment-inferred, and what happens if one of them stops shipping?
• Ask: 2 of Anthropic's tracked suppliers are a monopoly or leader (including CoreWeave, NVIDIA). Which of those are actually single-source for the process step you care about?
• Ask: The dataset names 2 tracked customers for Anthropic. How concentrated is demand among those names, and which of them have a documented second source in this graph?
• Ask: Anthropic is headquartered in United States. What jurisdiction, export-control or logistics risk sits at that HQ, versus at the fabs or plants the filings actually name?
• Ask: Anthropic is private in this dataset (no ticker). What primary-source filings or audited statements exist, and what is still only a one-liner?

Investor checklist

Investor section

• Ask: Is the graph position (monopoly / leader / major / challenger / niche) the thing you are underwriting, or a product line the one-liner does not cover?
• Ask: Which named customers in this graph could switch, and which named suppliers could not be replaced in a year? Count only names this dataset actually lists.
• Ask: What is not in this dataset — site-level capacity, contract terms, revenue split by customer — that you still need from a filing or a call?