In May 2024, America's biggest seafood chain filed Chapter 11 in Orlando with about 100,000 creditors. The autopsy — bankruptcy filings, court records, and a 2026 creditor lawsuit — reads like a graph query: an owner who was also the sole supplier and a creditor, and a landlord created by the chain's own leveraged buyout. Below is that story as an entity graph, built entirely from public sources. Click any circle to drill in.
None of this was hidden. Every fact in the graph above was public — SEC-adjacent filings, court records, trade press — before the collapse. What was missing was the join:
One entity, three roles. Thai Union was simultaneously Red Lobster's controlling owner, its shrimp supplier, and one of its creditors. In 2023, after a "quality review" that the 2026 creditor lawsuit calls pretextual, competing shrimp suppliers were dropped — leaving the owner as sole source for roughly half of all shrimp types, at prices filings describe as above market. Decisions steered that way cost the chain an estimated $76 million. A dependency graph flags an owner-supplier-creditor triangle automatically; a spreadsheet of vendors never will.
The landlord the company built for itself. Golden Gate Capital's 2014 buyout was financed by selling the land under ~500 restaurants for $1.5 billion — cash that went to the deal, while the leases (many above market) stayed with the restaurants. Final-year rent: over $190 million. In the graph, that's a single load-bearing edge pointing at a REIT.
The promo was a symptom, not the cause. $20 Ultimate Endless Shrimp lost $11 million — the number everyone laughed at. The structural dependencies above it are what actually broke, and they were visible years earlier to anyone — or any AI — that could query the graph.
Red Lobster's new CEO says he'll build the most AI-forward restaurant company in America. Below the chatbot layer, this is what AI-forward means: your suppliers, landlords, owners, and contracts as a scored graph your team — and your AI — can query before the risk becomes a filing. I build these in weeks, from your data and public records. The lab takes two builds a quarter.
Book a 30-minute review