AI Infrastructure Database

This site is a research database, not a news blog. Its core is a knowledge graph of the AI industrial ecosystem: chips, servers, networking, memory, foundries, packaging, cooling, data centers, power plants, clouds, software, models, products, companies, people, benchmarks, papers and datasets — joined by typed, bidirectional relationships, so that one click separates a GPU from its foundry, its memory supplier, the rack it ships in, the campus that deploys it and the power deal behind that campus.

It exists to answer the questions the AI boom keeps raising: What hardware powers modern AI? Who manufactures each component? How much electricity does it consume, and where will that power come from? Which public companies sit at each layer? What could bottleneck the next five years — and what do the people building it predict for 2030? Technical depth is available everywhere, but every page is written to be understandable by an educated non-engineer.

The data lives by strict rules: every figure is labeled with its basis (vendor spec, vendor claim, reported, estimate, official); every record carries an as-of date; undisclosed values say "not publicly disclosed" rather than guessing; forward-looking claims are attributed to whoever made them, with dates. News is aggregated from named mainstream and specialist outlets and always links to the original article; short data briefs written in-house with AI assistance are labeled as such. Nothing on this site is investment advice.

The architecture is built for growth: the same graph engine that holds today's records is designed to hold thousands more — new chips, campuses, models and relationships are added as records, not redesigns. Corrections and additions are welcome via the contact page.